Our selection of the top business news sources on the web.

AM edition. Issue number 1352

Latest 10 stories. Click the button for more.

Quote: Jeetu Patel - President and chief product officer at Cisco

"The amount of infrastructure needed for an agent is meaningfully higher than for a chatbot. For every human you might have 10, 100 or on the aggressive side 1 000 agents... They just keep working and that consumes a chunk of [compute]." - Jeetu Patel - President and chief product officer at Cisco

Enterprise AI is colliding with a hard economic constraint: systems that act on behalf of humans, rather than simply converse with them, generate continuous demand for compute, memory, storage, networking and security that scales non-linearly with adoption. What began as an experiment with a few chatbots embedded in customer support flows has evolved into fleets of autonomous or semi-autonomous agents woven into operational systems, provoking a structural rethink of infrastructure capacity planning, architecture and cost allocation. As organisations discover that a single knowledge worker can be surrounded by dozens or hundreds of always-on software entities, the initial enthusiasm for ubiquitous AI is running directly into power limits, data-centre constraints and budget ceilings.

From conversational interfaces to autonomous execution

The crucial distinction is architectural. A typical chatbot is a reactive interface: it receives a text or voice prompt, runs a model invocation, retrieves data from a knowledge base or simple API, and returns a response. It behaves largely as a query layer on top of existing systems, often operating in a single-turn or short multi-turn context without independent goals, long-running state or direct write access to critical applications. Infrastructure demand in this pattern is relatively bursty and predictable: each user query maps to a bounded unit of computation and network traffic, which can be statistically smoothed across a population of users.

Agents invert this relationship between interface and execution. An AI agent is typically defined as a goal-driven, autonomous system that can perceive context across multiple data sources, reason about what to do next, select tools, and execute multi-step workflows across business systems. Rather than stopping at an answer, agents act: they update records in line-of-business applications, orchestrate workflows in CRM or ITSM platforms, trigger network changes, or coordinate with other agents to complete composite tasks. The conversational surface, if it exists at all, becomes only one of many possible triggers. The core of the system is an execution loop that persists until the goal is completed or escalated.

This difference in behaviour drives a step change in infrastructure needs. A chatbot largely consumes compute at inference time, with minimal persistent state and shallow system integration. An agent, by contrast, is bound into the fabric of enterprise systems and may hold long-lived context, subscribe to event streams, maintain memories and embeddings, and repeatedly invoke models and tools until objectives are satisfied. That shift from "answer once" to "keep working" is what turns a marginal AI workload into a structural infrastructure commitment.

Why one human can imply hundreds of agents

The claim that a single human may be associated with 10, 100 or even 1 000 agents captures an emerging pattern in enterprise design. Knowledge workers are increasingly surrounded by specialised AI entities: a scheduling agent, a document drafting agent, a data-pull agent for analytics, multiple monitoring agents for infrastructure, and domain-specific agents embedded in each SaaS tool. Each of these has a narrow remit, but together they form an ecosystem of micro-work units continuously reacting to events, logs, tickets and user behaviour.

From an architectural perspective, this resembles the move from monolithic applications to microservices, but with an added dimension of autonomy and continuous reasoning. Instead of a single large conversational system, enterprises are decomposing tasks into networks of agents with defined roles and tool sets, sometimes coordinated by higher-level orchestrator agents. That design enables agility and specialisation, yet it multiplies the number of active components that must be powered, connected, secured and observed. Each agent may appear lightweight in isolation, but at the scale of tens of thousands of employees, the number of concurrently active entities can rise by two or three orders of magnitude.

Importantly, many of these agents are not user-facing at all. Infrastructure teams are experimenting with agents that monitor network telemetry, propose or implement configuration changes, execute test suites, and validate rollouts using digital twins of production networks. Security teams are piloting detection and response agents that ingest logs, enrich alerts, and autonomously contain incidents within pre-defined guardrails. In these environments, the mapping of "agents per human" reflects not only personal productivity tools but also operational automation embedded deep in the stack.

The continuous compute drain of agentic workloads

Because agents keep working in the background, infrastructure load is no longer driven solely by explicit user queries. Agents subscribe to message queues, listen to event buses, and schedule periodic checks; they may maintain rolling embeddings of new documents, update knowledge graphs, or re-evaluate risk scores as fresh data arrives. Each step involves model invocation, data movement, or both. Even if individual actions are small, the aggregate forms a constant baseline of demand that persists 24/7.

For cloud or on-premise operators, this means the cost function moves from primarily variable, usage-driven spend towards a blend of fixed and semi-fixed commitments. If represents the number of human users and the average number of agents per user, the potential number of agents is . Yet the relevant quantity for infrastructure sizing is the set of concurrently active agents, which depends on trigger frequencies, task durations, and coordination patterns. A naive but illustrative view is that if each agent has an activity duty cycle (fraction of time spent computing or transferring data), then expected concurrent load scales with . Even modest values of and can generate a substantial continuous baseline when is in the tens of thousands.

Network demand behaves similarly. Agentic systems that act across hybrid environments must traverse data-centre fabrics, campus networks and WAN links to reach telemetry sources, SaaS APIs and edge devices. Jeetu Patel has described how emerging agent workloads drive persistent high-volume east-west and north-south traffic, contributing to what he terms a network supercycle as enterprises upgrade switching, routing and optical capacity to cope with AI-driven data flows. These patterns differ markedly from classic web or batch analytics workloads, with more continuous, many-to-many flows and tight latency expectations for closed-loop control.

Agentic AI as an infrastructure business

The strategic implication for infrastructure providers is straightforward: agents monetise hardware. Every move from static chat interactions to autonomous workflows increases the mix of workloads that are both compute-intensive and long-lived, lengthening hardware refresh cycles and justifying investment in high-performance networking, accelerators and storage. Vendors positioned across the stack are racing to articulate platforms specifically designed for agentic operations, with unified views of networking, security, compute and observability that treat agents as first-class operational actors.

One expression of this is the emergence of platforms that explicitly support "humans and AI agents running critical infrastructure together" and provide a single control plane for both traditional systems and autonomous components. These platforms aim to normalise agents as operational peers: they authenticate, authorise and log agent actions; expose natural-language interfaces for defining new workflows; and integrate with existing observability stacks to track performance and anomalies in agent behaviour. This positioning reflects a belief that the long-term value in AI will accumulate less in standalone chat applications and more in integrated operational systems where agents co-manage infrastructure and business processes.

Economic tension: cost, usage and budget shock

For enterprises, the same dynamic poses uncomfortable questions about cost and governance. Early adoption of generative AI often focused on text chat interfaces, where spending could be bounded via rate limits, per-seat pricing and clear attribution of usage to teams. The transition to agents challenges these controls. An agent that can trigger actions in ITSM, CRM or ERP systems may also quietly trigger costs: more API calls to third-party platforms, more log ingestion into SIEM tools, higher storage for generated artefacts, and, crucially, higher inference and orchestration compute.

Reports of organisations pulling back on expansive AI deployments due to spiralling cloud bills capture this emerging reality. Once agents are embedded into daily operations, turning them off is not as simple as disabling a chatbot widget. They become entangled in workflows and SLAs. Finance teams, meanwhile, discover that AI line items are not merely "experimental" but have become semi-fixed operating costs. The ratio of spend to value becomes harder to measure when hundreds of agents operate in the background, each doing small, distributed pieces of work whose direct contribution to revenue or cost savings is difficult to isolate.

Vendors of foundation models and APIs have responded with tiered pricing, volume discounts and specialised tokens for specific use classes, but the basic arithmetic is unchanged: sustained autonomy consumes sustained resources. Internally, CIOs and CFOs are being forced to adopt more granular cost-allocation models, tracking which business units are responsible for which agent fleets, and tying deployment approvals to explicit ROI hypotheses in terms of reduced headcount, faster cycle times, or risk reduction.

Why agents are harder to host than chatbots

From an engineering standpoint, agentic systems impose stricter requirements across multiple dimensions of infrastructure. First, they demand more sophisticated state management and storage, as agents need to remember prior context, plan over long horizons, and coordinate with other entities. This often implies vector databases for embeddings, graph stores for relationships between entities, and durable logs for auditability. All of these add to storage and I/O requirements compared with a stateless chatbot backed by a simple knowledge base.

Secondly, agents need deeper integration with identity and access management. Because they can execute actions that affect real systems, they must be governed by policies defining which tools they can call, what data they can read or write, and under what approval conditions. This adds complexity to security architecture: agents require service identities, rotating credentials, fine-grained permissions and sometimes per-action human approvals, all of which must be enforced consistently across hybrid and multi-cloud environments.

Thirdly, the network and compute layers must be designed for low-latency, high-reliability closed loops. Agent workflows that modify infrastructure or process financial transactions cannot tolerate unpredictable delays or frequent timeouts. This drives demand for high-bandwidth, low-loss fabrics inside data centres, intelligent traffic engineering across WANs, and tight coupling between observability systems and control planes so that anomalies trigger automated mitigation rather than manual tickets. These characteristics go well beyond what is needed to serve sporadic chatbot traffic.

Power, sustainability and the physical edge of the agent boom

Behind these logical architecture concerns lies a physical constraint: power. Large-scale AI deployments are increasingly bottlenecked not by chips alone but by the availability and cost of electricity to feed data-centre clusters and edge compute nodes. When each human user implies potentially hundreds of always-on agents, total energy consumption can rise sharply, especially if models are large or poorly optimised. Leaders in the field have warned that planning for power is a first-order requirement for CIOs considering agentic AI at scale, rather than an after-thought once use cases have been defined.

This constraint interacts with geography and regulation. Regions with limited grid headroom or stringent environmental policies may face harder trade-offs between expanding AI capacity and meeting sustainability commitments. Enterprises are therefore exploring techniques to bend the resource curve: model distillation and quantisation to reduce inference cost; adaptive scheduling that defers non-urgent agent activity to off-peak times; on-device or near-edge inference for local tasks; and more precise scoping of agent authority so that they do not perform unnecessary or redundant work.

Security and the need for agent-aware defences

Security concerns compound the infrastructure challenge. Agents that autonomously operate across networks and applications introduce new attack surfaces and failure modes. If compromised, an agent with write access to critical systems could cause damage at machine speed. Even without malicious interference, mis-aligned objectives or prompt injection attacks can lead agents to take unintended actions. Security architectures are being retooled to treat agents as high-value, high-risk entities that must be monitored, constrained and protected.

Several patterns are emerging here. One is the fusion of security controls directly into networking and AI infrastructure, so that traffic to and from agents can be inspected, segmented and policy-controlled without relying solely on application-level safeguards. Another is the deployment of "security agents" that watch other agents, analysing behaviour for anomalies, enforcing guardrails and escalating suspicious activity for human review. This meta-agent layer, however, introduces yet more continuous workload, reinforcing the original observation that agentic ecosystems increase the total infrastructure footprint.

Debates and objections: are hundreds of agents per human inevitable?

There is a live debate about whether the projected density of agents is a necessary outcome or an artefact of early design choices. Critics argue that current enthusiasm for fine-grained agents may be over-engineering: instead of dozens of micro-agents per user, organisations could converge on a smaller number of more capable, multi-role agents, reducing orchestration overhead and infrastructure load. Others suggest that existing automation and rules-based workflows can absorb a substantial share of tasks without requiring full agent autonomy, using agents only at decision points where judgment and flexible reasoning are genuinely required.

There are also concerns that agent proliferation could outpace human ability to understand and govern system behaviour, leading to opaque networks of interacting entities whose collective impact is difficult to predict. In response, some practitioners advocate for stricter criteria to qualify a component as an agent: it must demonstrate clear, measurable business value, operate within narrow and auditable boundaries, and be subject to regular decommissioning reviews to prevent uncontrolled growth in the agent population.

Yet even more conservative designs acknowledge that the direction of travel is towards greater autonomy and deeper integration of AI into operational systems. Once organisations experience the productivity and resilience benefits of agents that can, for example, validate network changes against digital twins before deployment, or automatically draft and triage IT tickets, the pressure builds to expand their remit. The resulting increase in sustained compute and network demand may be somewhat optimisable, but not easily reversed.

Why the distinction matters for strategy and policy

Understanding the gap between a chatbot and an agent is not a matter of terminology; it is a strategic planning issue for boards, regulators and investors. Boards need to grasp that AI initiatives framed as "assistive chat" can evolve into critical operational dependencies with recurring infrastructure costs and systemic risk profiles. Regulators, particularly in sectors such as finance, health and critical infrastructure, must recognise that autonomous systems executing actions require different oversight, audit trails and safety cases compared with systems that merely answer questions.

Investors and market analysts, meanwhile, are watching how infrastructure vendors position themselves around this transition. Companies with the ability to supply not just model capabilities but also the secure, power-efficient, network-rich environments in which agent fleets can safely operate may enjoy durable demand, as agentic workloads lock in customers for multi-year infrastructure refresh cycles. Conversely, enterprises that underestimate the infrastructure implications of moving from chatbots to agents may find early AI gains eroded by escalating costs and operational fragility.

The deeper story is that the move from conversational AI to agentic AI transforms AI from an application feature into an organising principle for enterprise architecture. Once software entities can act with sustained autonomy, the infrastructure beneath them becomes a strategic asset and a potential bottleneck. The observation that supporting this shift requires markedly more infrastructure than hosting simple chatbots is less a prediction than a description of a redesign already under way in large organisations worldwide.

Term: Reinvestment rate - Corporate finance

"The reinvestment rate refers either to the percentage of earnings a company retains to fund capital expenditures and working capital for future growth, or to the assumed interest rate at which project cash flows are reinvested over time. As a metric, it determines a firm's fundamental growth rate when multiplied by its return on invested capital (ROIC)." - Reinvestment rate - Corporate finance

Corporate growth is ultimately constrained by two linked forces: the amount of cash a firm chooses to plough back into the business, and the return it can earn on that incremental capital. When either element is weak, growth stalls or destroys value; when both are strong and sustained, operating income, intrinsic value and, eventually, share prices can compound for decades. The practical challenge in corporate finance is to understand how much to reinvest, where to reinvest it, and what growth rate those choices imply.

The underlying economic mechanism

Every operating business generates cash from its core activities and then faces a recurring allocation problem. Management can distribute cash to investors, leave it idle, or recycle it into new projects, capacity, and working capital. The portion recycled determines the scale of new assets being created; the economic quality of those assets is captured by their return on invested capital. Together, these two levers drive how quickly operating earnings and enterprise value expand over time.

If a firm channels a large share of its after-tax operating profits into capital expenditure and working capital, but those investments only earn a low return, the result is hollow growth: the balance sheet and revenues expand, yet value barely increases or even declines. By contrast, a firm that reinvests modestly at exceptionally high returns can grow value rapidly, even if headline revenue growth appears moderate. The reinvestment decision is therefore inseparable from the expected return on incremental capital.

Two core meanings: retained earnings and project cash flows

In corporate finance practice, the reinvestment rate appears in two related but distinct contexts.

First, at the firm level, it often describes the share of earnings retained instead of being paid out as dividends. In this framing, one can express a simple reinvestment rate as retained earnings divided by net income, or equivalently one minus the dividend payout ratio. A company that pays out 40 % of its income as dividends implicitly reinvests 60 % back into the business. This perspective is common in discussions of sustainable growth, where analysts link the retention ratio to return on equity.

Second, in valuation and operating modelling, practitioners use a more granular notion tied directly to operating cash flows. Here, reinvestment is measured as the net amount spent on long-lived operating assets and additional working capital to support future operations. The focus is not on accounting earnings per se, but on how much of the after-tax operating profit is redirected into expanding the capital base.

Both perspectives are analytically similar: they describe the share of internally generated resources that is not distributed and is instead committed to new or expanded operations. The difference lies in the accounting definitions used and the level at which capital allocation is examined.

Practical measurement from financial statements

To operationalise the concept for valuation and performance analysis, most practitioners work with after-tax operating income and the cash invested in operating assets. A widely used definition takes the form:

where:

- , representing gross capital expenditure minus the portion simply replacing worn-out assets.

- is the change in net working capital, usually defined as non-cash current assets minus non-interest-bearing current liabilities.

- is net operating profit after tax, typically calculated as .

This formulation links the reinvestment rate directly to the cash needs of the business. Net capital expenditure captures incremental investment in property, plant, equipment and other long-term operating assets. The working capital component reflects the financing required to hold inventories, extend receivables, and support day-to-day operations as the business grows.

By dividing the total reinvestment by NOPAT, the metric expresses what fraction of the current after-tax operating income is being committed to future growth, rather than being available for distribution to equity and debt holders. A reinvestment rate of 30 % in this sense means that 30 % of the firm7s operating earnings are being recycled into the business each period.

Connection to ROIC and fundamental growth

The importance of this ratio becomes clear when it is coupled with the return on invested capital. ROIC itself is generally defined as:

where invested capital comprises the operating assets funded by long-term providers of finance, often approximated as interest-bearing debt plus equity minus non-operating cash.

Under the simplifying assumption that reinvested funds earn the existing ROIC, the expected growth in operating income can be expressed as:

This relationship mirrors the classic sustainable growth identity that uses the retention ratio and return on equity, but focuses instead on operating income and the total capital base. It says that if a firm retains a certain share of its operating earnings and earns a given percentage return on the resulting incremental capital, its operating income will grow at roughly the product of these two terms. For example, a company reinvesting 40 % of its NOPAT at a 15 % ROIC would be expected to grow operating income by approximately , or 6 % per year, absent major structural shifts.

Investors and valuers often use this formulation as a bridge between historical financials and forward-looking valuation models. Once a stable reinvestment rate and ROIC are estimated, one can derive a baseline growth assumption for operating cash flows and thus for discounted cash flow models.

Intrinsic value compounding and incremental returns

From a valuation standpoint, what truly matters is the return on incremental invested capital (ROIIC), not merely the aggregate historical ROIC. If the incremental projects funded by reinvestment earn lower returns than the existing asset base, the headline ROIC may remain high for a while, even as value creation deteriorates. Conversely, if new investments can be made at similar or higher returns, the enterprise7s intrinsic value can compound at a rate close to reinvestment rate times ROIIC.

Analysts therefore pay close attention to whether a company can expand without diluting its return profile. A business that reinvests 50 % of earnings at a 20 % ROIIC can grow intrinsic value at roughly 10 % annually, assuming the economics are sustainable. Once the opportunity set shrinks and new projects fall closer to the cost of capital, reinvestment creates far less value and may even destroy it if ROIIC drops below the weighted average cost of capital.

This interplay generates a central tension in capital allocation: cutting reinvestment boosts near-term free cash flow and dividends, but may slow the compounding of intrinsic value; raising reinvestment can accelerate growth, but only if incremental returns remain adequately high. Over time, markets tend to reward firms that maintain a disciplined balance between these forces.

Alternative reinvestment definitions and payout links

In some contexts, a simpler, earnings-based reinvestment metric is used:

This definition aligns closely with equity analysis that focuses on the growth of book equity and earnings per share. If return on equity remains stable, the sustainable growth in earnings can be approximated as retention ratio times ROE, analogous to the ROIC framework:

However, because it is tied to net income rather than operating income and ignores debt-funded investment, this formulation is less informative about the total economic reinvestment in the business. It is therefore better suited to analysing shareholder payout policy than to modelling operating growth and enterprise value.

Reinvestment rate as assumed project reinvestment yield

Beyond firm-level capital allocation, the term also appears in project appraisal as the assumed rate at which interim cash flows are reinvested. In internal rate of return (IRR) calculations, for example, it is often implicitly assumed that intermediate project cash flows can be reinvested at the IRR itself. Some analysts regard this assumption as unrealistic and instead prefer to use a more conservative reinvestment rate, linked to the cost of capital or to observable market yields.

In modified internal rate of return (MIRR) frameworks, one explicitly specifies a reinvestment rate at which project cash inflows are compounded until the end of the project. This reinvestment rate need not match the project7s own IRR; it often reflects the firm7s opportunity cost of capital or another internally available rate. Changing this assumed rate can materially alter the MIRR and thus the apparent attractiveness of a project, highlighting the sensitivity of appraisal metrics to reinvestment assumptions.

Parameter meanings and estimation challenges

Translating the theoretical relationships into usable estimates requires careful parameter choices.

- NOPAT is meant to capture after-tax operating performance independent of financing decisions. Estimating it may involve normalising margins, adjusting for non-recurring items, or reclassifying certain expenses as capital items when they create long-lived benefits (for example, some research and development spending).

- Net Capex must distinguish maintenance investment, which simply preserves existing capacity, from growth investment, which expands it. While the basic formula subtracts depreciation from gross capex, in practice analysts sometimes adjust this further, particularly for businesses with lumpy investment cycles or significant intangible expenditures.

- Working capital swings can distort single-period measures. A temporary build-up of inventory or a deliberate change in credit terms may cause to spike, making the reinvestment rate appear unusually high or low. Averaging over several years can provide a more stable picture.

- ROIC itself can be computed using beginning-of-period, end-of-period, or average invested capital, and may require adjustments for non-operating assets, goodwill, and capitalised operating leases. These choices affect the measured level and trend of returns.

Because these parameters are all interdependent, a mechanistic use of the formulas can be misleading. A robust analysis cross-checks implied growth rates against observed revenue trends, market saturation, competitive dynamics, and management guidance.

Major schools of thought on reinvestment policy

Corporate finance theory offers several perspectives on the optimal reinvestment rate.

One school emphasises a value maximisation rule: firms should reinvest only when the expected return on capital exceeds the cost of capital, and distribute any surplus cash. Under this view, excessively high reinvestment rates in low-return projects represent agency problems or empire building. Strong ROIC coupled with disciplined, selective reinvestment is seen as the hallmark of effective management.

A second school highlights the strategic benefit of scale and market share. It argues that reinvesting heavily to build network effects, brand strength or cost leadership can justify temporarily depressed returns, as long as eventual ROIC on the expanded capital base exceeds the cost of capital. This approach is common in high-growth technology and platform businesses, where management may intentionally accept near-term low or negative accounting returns in pursuit of long-term competitive advantage.

A third perspective focuses on shareholder preferences. Income-oriented investors may favour lower reinvestment rates and higher payouts, while long-term growth investors may prefer aggressive reinvestment at attractive returns. In practice, boards attempt to align reinvestment and payout policies with the shareholder base they wish to attract.

Tensions and debates

Several recurring debates revolve around the reinvestment rate concept.

First, there is the question of profitability thresholds. The simple growth identity suggests that higher reinvestment always raises growth. Yet if ROIC falls below the cost of capital, faster growth can destroy value. Some commentators therefore stress that the reinvestment rate only contributes to value when incremental returns are positive and, crucially, exceed the hurdle rate. When the core business is structurally unprofitable, reinvestment simply scales up value destruction.

Second, analysts disagree on how quickly ROIC tends to revert as firms grow. Proponents of structural competitive advantage argue that certain businesses can sustain high ROIC for long periods, justifying high reinvestment rates. Others point to competitive entry and innovation pressures that push returns down over time, implying that reinvestment opportunities at attractive returns will be gradually exhausted. The truth varies by industry and firm, making empirical analysis essential.

Third, the measurement of reinvestment is increasingly complicated by the rise of intangible capital. Expenditures on software development, data assets, brand, and human capital may be expensed under accounting rules but function economically like capital investments. If these are not capitalised in analytical models, reinvestment rates and ROIC can be overstated, particularly for digital and service businesses. This has sparked ongoing efforts to adjust financial statements to better reflect economic reinvestment.

Why reinvestment rate still matters in modern corporate finance

Despite evolving business models and accounting complexities, the reinvestment rate remains central to understanding long-term value creation.

For investors, it offers a disciplined way to think about growth. Instead of extrapolating revenue expansions on the basis of narratives alone, analysts can ask how much capital will be required to support that growth and what returns it is likely to earn. A company promising 15 % annual growth but reinvesting only 10 % of NOPAT at a 20 % ROIC faces a mechanical inconsistency: the implied fundamental growth is closer to 2 % than 15 % unless leverage, margins, or asset turnover change substantially.

For managers, tracking reinvestment rates by business line illuminates where capital is genuinely productive. Units that absorb significant capital but fail to deliver corresponding NOPAT growth may need restructuring, divestment, or a change in strategy. Conversely, high-ROIC, capital-light segments might justify additional investment or acquisitions to scale their economics.

For boards and capital allocation committees, the reinvestment rate is a governance tool. It clarifies the trade-off between buybacks, dividends, debt reduction, and internal projects. A board that understands the firm7s opportunity set and ROIC trajectory can set target reinvestment ranges that maximise long-run value while maintaining financial resilience.

Finally, for valuation and risk management, the linkage between reinvestment rate, ROIC and growth provides a coherent framework for scenario analysis. Shocks to demand, changes in competitive intensity, or regulatory interventions can be translated into adjustments in reinvestment capacity and incremental returns, yielding revised growth paths and valuations.

In all these applications, the reinvestment rate serves not as an isolated ratio, but as one half of a dynamic pair with return on capital. Observed together through time, these metrics tell a story about how a firm converts today7s cash flows into tomorrow7s earning power. That story, more than headline earnings or short-term share price moves, lies at the heart of long-term corporate finance analysis.

Quote: Carter Busse - Workato chief information officer

"Our spend went up 7x the first day [when Anthropic switched it over to token-based pricing in May] and I'm like, oh sh*t, we created a monster. [Large language model] companies have been subsidising all of our usage and now no longer. User-based pricing shelters you." - Carter Busse - Workato chief information officer

Enterprise adoption of generative AI has exposed a structural tension between enthusiasm for ubiquitous assistants and the brute economics of large language model computation. Organisations raced to roll out AI copilots to every knowledge worker, often under seemingly generous, user-based licensing that made usage feel close to free at the margin. That changed abruptly once providers shifted to pricing rooted in the true cost driver: tokens processed per request. The sudden visibility of usage-based bills has forced chief information officers to confront whether they have built a durable productivity platform or an uncontrolled cost engine running on someone else's balance sheet.

From enthusiasm to sticker shock

Workato offers a stark example of this pivot. As an integration and automation platform, it was a natural early adopter of agentic AI across internal workflows and customer-facing automation. Once generative tools were rolled out widely, adoption surged: Workato reported more than tenfold growth in LLM usage and a material revenue uplift tied to AI-powered features. For a time, costs were muted because major model vendors, including Anthropic, relied on user- or seat-based pricing and flat enterprise arrangements that decoupled consumption from marginal spend. The economic signal reaching the CIO was weak: expanding access felt low-risk because the incremental cost of another prompt, conversation, or agent run was effectively zero to the business.

The turning point came when Anthropic moved its enterprise customers onto token-based billing for Claude usage. Instead of paying primarily per user, Workato began paying directly for the volume of tokens consumed across all its internal and product workloads. Overnight, cost visibility flipped. A pattern of generous experimentation, long conversations, and proliferating internal tools translated into a first-day cost that was roughly seven times prior levels, revealing just how much latent demand had been masked by the previous pricing model. What had looked like measured adoption was, financially, a rapidly scaling compute obligation that had not been governed as such.

How user-based pricing acts as a de facto subsidy

User-based pricing can be thought of as a coarse-grained hedge against the variability of generative workloads. Under a flat per-seat model, a provider implicitly averages heavy and light users, leaving the supplier to absorb volatility in usage and peak demand. For enterprise buyers, the value proposition is predictability: once licences are purchased, finance teams can forecast AI costs with the same tools they use for other SaaS budgets, regardless of whether an individual employee sends ten prompts per month or ten thousand.

This arrangement amounts to a cross-subsidy. Heavy users and intensive automation workloads consume far more underlying tokens and compute than light users, but are billed identically so long as they fit under the plan's qualitative usage limits. The supplier is effectively underwriting the risk that a subset of customers exploit the flat pricing to build high-duty cycles, long-context workloads, or agent frameworks that keep models running continuously. For a while, competitive dynamics encouraged this behaviour: vendors prioritised adoption and growth metrics, accepting that early-stage monetisation might lag behind actual compute costs.

Once models became central to core business processes and total token volumes began to soar, the imbalance became untenable. Shifting from seat-based to token-based billing is the supplier's way of converting an averaged, opaque cost structure into one where revenue tracks the primary cost driver: the number and type of tokens processed. Instead of subsidising heavy users through broad user categories, providers charge each organisation in line with its actual compute footprint.

The mechanics of token-based pricing shocks

Token-based billing operates on a simple arithmetic relationship: total bill equals token volume multiplied by the rate per token, differentiated between input and output. In its current enterprise API pricing, Anthropic charges separate rates for input tokens - the text, documents, and context supplied to Claude - and output tokens - the model's generated responses. For many flagship models, published rates cluster around to dollars per million input tokens and to dollars per million output tokens, with more advanced models commanding higher prices. Other vendors such as OpenAI and Google follow broadly similar structures, though with widely varying rates across model tiers.

The shock arises because enterprise buyers often underestimate both the volume of tokens and how quickly compounding factors magnify usage. First, conversational use encourages verbosity. Users ask broad questions, paste large documents, and accept multi-paragraph answers. Each interaction consumes both input and output tokens, and for many models, output tokens are priced several times higher than input tokens. Second, long-context capabilities enable prompts that include extensive histories, knowledge bases, or email threads. Once the context window stretches into hundreds of thousands or even a million tokens, a single request can carry a cost multiple orders of magnitude larger than a simple chat, especially if premium modes for long context or fast inference are triggered.

Third, agentic workflows - a particular focus for Workato - chain multiple model calls together. An agent tasked with, say, triaging an IT ticket will interpret the request, query knowledge bases, draft responses, perhaps call downstream tools, and refine recommendations, each step incurring additional tokens. Where a human sees one business action - resolve a ticket - the billing system sees a series of separate model invocations. If this pattern is replicated thousands of times per day across customer support, sales operations, and back-office processes, total monthly token usage can explode without any single interaction appearing excessive.

Under user-based pricing, these dynamics were effectively invisible. Under token-based billing, they manifest instantly in the invoice. A sevenfold jump in spend on day one is a symptom of previously hidden intensity, not a sudden behavioural change. The organisation did not dramatically alter how it used AI; it simply began paying in line with the true cost structure underpinning its workloads.

Why providers must move away from implicit subsidies

On the supplier side, there are structural reasons why AI companies can no longer sustain broad, implicit subsidies at scale. Training and serving frontier models require massive investment in specialised hardware, energy, and engineering. Reports suggest that revenues at leading LLM firms now run into tens of billions of dollars annually, but those revenues are tightly coupled to equally significant capital and operating expenditures on GPUs and data centre infrastructure. If enterprise customers consume millions or billions of tokens under flat-price contracts, the provider bears the risk that actual compute costs exceed the effective revenue per token.

Moreover, token pricing has become a competitive battleground. In 2026, published API prices span a range of roughly six hundred times between the cheapest small models and the most advanced reasoning systems. Some entrants aggressively discount to gain share, while incumbents experiment with premium tiers for speed, long context, or jurisdiction-specific inference. Maintaining cross-subsidies under these conditions becomes strategically dangerous: it obscures whether a model's economics are genuinely sustainable or propped up by temporarily cheap capital and investor tolerance for losses.

Switching to usage-based billing restores economic discipline. Revenue becomes a near-linear function of tokens processed, allowing capacity planning, data centre investment, and R&D schedules to be benchmarked against projected token volumes rather than abstract user counts. It also creates room for finely tuned price discrimination: different rates for input versus output, surcharges for fast modes or extended context, regional multipliers, and discounts for batch processing or caching, all of which Anthropic and its peers now deploy at scale.

User-based pricing as a psychological and governance shield

From the enterprise perspective, per-user pricing offers more than just financial predictability; it provides a psychological and operational shield that encourages experimentation. Employees are more likely to integrate AI into daily work when they know they are not triggering metered charges with every prompt. Citizen developers within a platform such as Workato can prototype agentic workflows, automate routine tasks, and iterate on internal tools without negotiating budget allocations for each new integration. The absence of visible marginal cost fosters the kind of bottom-up innovation that many digital leaders seek to cultivate.

However, that same shelter can delay the establishment of governance mechanisms commensurate with the technology's power. When usage feels free, few teams invest in monitoring token consumption, optimising prompt length, or choosing models appropriate to each workload. Security and compliance reviews might focus on data handling and hallucination risk, while financial controls lag behind. In such an environment, the shift to token-based billing functions like a sudden exposure of hidden leverage: what looked like a manageable pilot proves to be a complex portfolio of high-throughput workloads with no cost controls.

Global Advisors News Brief - June 20 2026

Headlines for the last 24hrs

1. The AI Talent War Escalates as Top Researchers Shift Alliances Between Tech Giants

2. Rising AI Infrastructure Costs Force Tech Giants to Test Their Pricing Power

3. Geopolitical Volatility and US-Iran Nuclear Diplomacy Impact Global Financial Markets

4. Re-evaluating Enterprise AI Strategy: Moving Beyond the Hype to Practical Integration

5. Monetary Policy Shifts and High Interest Rates Squeeze Corporate Debt Strategies

6. Supply Chain Bottlenecks and Inflation Squeeze Food and Consumer Goods Sectors

7. Regulatory Scrutiny and Geopolitical Concerns Create Hurdles for Mega-Mergers

8. Government-Influenced Tech Alliances Reshape Domestic Semiconductor Supply Chains

9. Grid Modernization and Climate Philanthropy Accelerate Clean Energy Infrastructure

10. Private Equity Adapts to Tight Markets with New Mid-Market Exit Blueprints

Time window: 2026-06-19T18:35:19.862Z to 2026-06-20T18:35:19.862Z

Term: 'Capital Maintenance Rate' or 'Capital Recovery' - Financial accounting

"The 'Capital Maintenance Rate' or 'Capital Recovery' is a specialised term used primarily in corporate accounting and regulated industries. It refers to either the rate at which an entity preserves its original capital base over time or, in a regulatory context, the discount rate used to calculate future returns that ensure investors recover their initial investment." - 'Capital Maintenance Rate' or 'Capital Recovery' - Financial accounting

Profit figures are only meaningful if the underlying capital base has been preserved. Without a clear notion of what must be kept intact before gains are recognised, reported earnings can conceal erosion of the business foundation, distort valuations, and misalign incentives between managers, investors, and regulators. The idea that returns should be assessed only after ensuring that the original capital has been maintained runs through financial accounting, regulatory pricing, and investment analysis, even though it is expressed through different mechanisms in each setting.

Economic substance: preserving the investment before counting profit

The economic concern is straightforward: investors commit funds and want assurance that reported income does not simply reflect running down the assets they have already paid for. Capital maintenance, sometimes called capital recovery, insists that profit exists only after the entity has first restored its capital to an agreed baseline level. Until this baseline is met, any apparent surplus is treated as recovery of the initial investment rather than genuine gain.

In corporate accounting, this principle leads to the rule that income is recognised only after the costs of operations during the period have been fully recouped and the capital at period end is at least as high as at period start, adjusting for contributions from and distributions to owners. In regulated industries such as utilities, the same logic is applied through an allowed rate of return or discount rate designed to ensure investors recover their initial outlay, typically via tariffs or regulated prices that generate sufficient future cash flows.

Seen this way, any notion of a "capital maintenance rate" or "capital recovery rate" is a way of translating the preservation requirement into a percentage per year that must be achieved, either in accounting terms (so that capital at the end date matches the economic value at the beginning) or in regulatory pricing (so that discounted cash flows equal the invested base).

Financial capital maintenance in accounting

Within conceptual accounting frameworks, financial capital maintenance defines profit as the increase in net assets over a period after excluding owner transactions, provided that the financial capital at the end of the period is not lower than at the beginning. Under this view, capital is measured as the monetary value of net assets.

Let be the value of net assets at the beginning of the period, and the value at the end, both excluding owner contributions and distributions. Profit under financial capital maintenance can be written as:

However, this simple expression hides a crucial choice: whether and are measured in nominal monetary units or in units of constant purchasing power. Two main variants arise:

- Financial capital maintenance in nominal units: capital is maintained if the nominal currency amount of net assets is preserved. Profit is any nominal increase.

- Financial capital maintenance in constant purchasing power: capital is maintained only if the real (inflation-adjusted) value of net assets is preserved. Profit is any real increase beyond inflation.

In an inflationary environment, the difference is substantial. Under the nominal view, inflation-driven increases in asset values may be treated as profit even if the entity is simply maintaining its real capital. Under the constant purchasing power view, those increases merely preserve capital, and only gains beyond inflation are recognised as profit.

Where a capital maintenance rate is discussed in this context, it is often an implicit real or nominal rate of return required to keep at the targeted level after adjusting for inflation, revaluations, and retained earnings.

Physical capital maintenance and operating capacity

Physical capital maintenance interprets capital as the entity's productive capacity rather than the monetary value of net assets. Profit arises only if the entity's ability to generate goods or services (its operating capacity) at the end of the period exceeds that at the beginning, again excluding owner transactions.

Formally, if is the level of productive capacity at the start and at the end, profit exists only when:

Maintenance of physical capital means ensuring is at least ; any additional capacity is then seen as "profit" in physical terms. Depreciation policies, maintenance expenditure, and reinvestment decisions all feed into whether productive capacity is maintained or eroded. In industries with heavy equipment, this logic is echoed in maintenance capital expenditures: spending needed just to keep current operations and capacity intact rather than to expand them.

The choice between financial and physical capital maintenance frameworks affects not only reported profit but also perceptions of performance. A company may appear profitable in financial terms while its physical operating capacity is deteriorating because it under-invests in maintenance and renewal. Conversely, heavy reinvestment to preserve physical capacity could reduce reported profit in the short term while strengthening long-term viability.

Capital recovery as an investment and regulatory concept

In investment and regulatory analysis, capital recovery focuses on ensuring that the present value of cash flows generated by an asset or business equals at least the initial capital cost. The key object is a discount rate that ensures investors can recover their principal over time, after accounting for risk and the time value of money. This is often what practitioners mean by a capital recovery rate.

Suppose an investment costs today and generates expected cash flows over periods. A capital recovery rate is a discount rate that satisfies:

At this rate, the net present value of the investment is zero: the investor recovers exactly the original capital in present value terms, but no more. Any actual rate of return above implies positive value creation; any rate below implies value destruction.

In regulated sectors such as utilities and infrastructure, regulators often set allowed revenues or tariffs so that, over the life of the asset, the firm recovers its regulatory asset base plus a return that compensates for risk. Conceptually, the allowed rate of return plays the role of . The underlying logic is capital maintenance for investors: so long as the regulated business can earn at least that allowed return, investors recover their capital and are not worse off for having invested in an essential, regulated service.

Connecting accounting capital maintenance and capital recovery rates

Although the language differs, both accounting capital maintenance and capital recovery rate analysis revolve around distinguishing between recovery of invested capital and genuine surplus.

- In accounting, the focus is on defining profit as the amount by which net assets (financial view) or productive capacity (physical view) exceed the maintained capital base at the end of the period.

- In capital budgeting and regulation, the focus is on finding a discount rate at which the present value of cash flows equals the invested capital, separating recovery of principal from economic gain.

One can think of a capital maintenance rate in accounting as the minimum return that keeps the capital base intact given inflation, asset wear, and required reinvestment. If the entity earns exactly that rate, (or ) equals the required capital baseline, and reported profit (in a strict capital maintenance sense) is zero. Earnings above this rate represent profit; earnings below it indicate capital erosion.

Parameter meanings and related measures

When capital recovery is expressed in formulae, several parameters typically appear:

- : initial capital cost or investment outlay.

- : net cash flow in period , including operating cash flows and salvage values.

- : economic life of the asset or project.

- : discount rate that equates the present value of cash flows to ; the capital recovery rate.

Where inflation is material, analysts may distinguish between nominal and real versions of , aligning them with the financial capital maintenance variant chosen. If capital is defined in real terms, the relevant capital maintenance rate should be a real rate, with inflation handled separately.

In corporate planning, another related concept is the annual capital charge required to maintain capital. If is recovered over years at rate , the constant annual charge that recovers the investment can be written using the standard annuity factor:

This annual charge can be interpreted as the combination of economic depreciation and required return on capital. In cost-plus regulation and project appraisal, such a charge often underpins pricing decisions that respect capital maintenance.

Maintenance expenditure, capital erosion, and practical indicators

In asset-intensive businesses, preserving capital is not merely an accounting convention; it is an operational challenge. If maintenance spending is consistently below the level needed to offset wear, obsolescence, and safety requirements, the physical capital base shrinks, even if financial statements report profits. Practitioners sometimes use rules of thumb, such as flagging trouble when annual maintenance costs exceed a certain percentage of replacement asset value, or when deferred renewal backlogs reach certain indices.

For example, facilities managers may track the ratio of deferred renewal to current plant value as a condition index, and then set annual funding targets that at least maintain this index. The implied rate of renewal funding relative to the asset base is effectively a capital maintenance rate: if spending falls below this benchmark, the physical condition of the portfolio deteriorates; if it meets or exceeds it, capital is maintained or improved.

Similarly, investors estimating maintenance capital expenditure often distinguish it from growth capital expenditure, allocating only the former to maintaining existing capacity. If free cash flow is measured after maintenance capital but before growth capital, it becomes a closer proxy for the cash that remains after capital maintenance has been provided for. Here, the implicit capital maintenance rate is reflected in the ratio of maintenance capex to the asset base or to depreciation.

Schools of thought and conceptual debates

Capital maintenance has long been a contested area in accounting theory. Several tensions drive the debate:

- Nominal versus real capital maintenance: Advocates of nominal monetary measurement argue for simplicity and comparability, while critics highlight that inflation can make nominal profit figures misleading when capital is not preserved in real terms.

- Financial versus physical capital perspectives: Financial capital maintenance aligns naturally with investors focused on wealth measured in money terms, whereas physical capital maintenance emphasises the entity's productive capacity and is often favoured in sectors where capacity, service levels, and long asset lives dominate concerns.

- Historical cost versus current value: Maintaining capital measured at historical cost may be easier operationally, but it may not reflect the economic resources required to maintain service potential. Current value approaches (such as replacement cost) better capture what it would cost to restore capacity, but introduce volatility and valuation subjectivity.

- The role of price changes: Some frameworks treat increases in asset prices due to market movements as gains, while others classify them as capital maintenance adjustments rather than distributable profit, on the grounds that these increases are needed to maintain the capital base at current values.

These debates have direct implications for dividend policy, leverage decisions, and regulatory determinations of allowable returns. If capital maintenance is defined conservatively, fewer funds are treated as distributable profit, strengthening balance sheets but potentially reducing short-term shareholder payouts. If capital maintenance is defined loosely, distributions can be made while the real capital base quietly erodes.

Why the concept still matters

Even in an era of fair value accounting, complex financial instruments, and forward-looking valuation models, capital maintenance and capital recovery remain central to financial discipline for several reasons.

First, they impose a minimum standard on what counts as success. A business that generates accounting earnings but fails to maintain the economic value or productive capacity of its capital is not creating sustainable value. Investors and creditors increasingly scrutinise measures such as maintenance capex, renewal funding, and asset condition indices precisely to distinguish genuine value creation from disguised capital consumption.

Second, regulatory frameworks that determine allowed returns for infrastructure, utilities, and other essential services rely heavily on capital recovery logic. Tariffs and price controls are often calibrated so that, over the life of regulated assets, investors recover their capital plus a fair return, but no more. A mis-specified capital recovery rate can either over-burden consumers or deter investment in critical infrastructure.

Third, in inflationary or volatile environments, the distinction between nominal and real capital maintenance becomes more pressing. Firms that distribute nominal profits without reserving enough to preserve the real value of their capital can find themselves under-invested just when renewal is most expensive. Accounting frameworks that explicitly identify capital maintenance adjustments provide clearer signals to boards and investors about how much of reported income is safe to distribute.

Finally, the idea shapes internal performance measurement. By charging business units for the cost of capital and treating only returns above the capital recovery rate as economic profit, organisations align managerial incentives with the preservation and enhancement of the capital base rather than mere volume growth. This internal capital maintenance rate may be tied to the firm's weighted average cost of capital, adjusted for risk and inflation expectations.

Practical interpretation for analysts and practitioners

Analysts dealing with financial statements, valuation models, or regulatory determinations can use the underlying logic of capital maintenance and capital recovery in several practical ways:

- Interpreting profit: give more weight to earnings measured after maintenance capital expenditure and inflation adjustments, using them as indicators of surplus beyond capital preservation.

- Testing sustainability: compare maintenance-related spending and renewal indicators with depreciation and asset values to gauge whether capital is being maintained, consumed, or augmented.

- Calibrating discount rates: in project appraisal or regulatory contexts, identify the capital recovery rate as the threshold at which investors merely preserve capital, and then assess whether proposed returns provide an adequate margin above that threshold given risk.

- Designing payout policies: align dividend and buyback decisions with measures of profit calculated after capital maintenance, reducing the risk of over-distribution that weakens long-term capacity.

Seen across these domains, the ideas of capital maintenance and capital recovery act as a safeguard: they separate the return of capital from the return on capital. For anyone interpreting financial accounts, designing regulation, or evaluating investments, being explicit about the capital maintenance rate in play is essential to judging whether reported returns truly reflect value creation rather than the gradual depletion of the capital base that underpins future performance.

Quote: Vinod Khosla - Experienced technology investor

"I'm certain AI will do 80 per cent of the economically valuable work humans do today, for 80 per cent of all jobs, faster than most believe. The question isn't whether mass underemployment arrives by the next decade, but whether we have a coherent policy framework ready when it does." - Vinod Khosla - Experienced technology investor

Mass underemployment driven by automation is no longer a fringe scenario but a serious macroeconomic risk windowed into the next decade, not the next century. The central tension is brutally simple: digital systems that learn are scaling faster than labour markets, education systems and tax architectures can adapt. Productivity, deflation and abundant digital services sit on one side; concentrated capital ownership, job displacement and fiscal stress sit on the other. How governments, firms and citizens navigate this gap between technological tempo and institutional inertia will shape whether the transition feels like a ladder up or a trapdoor down.

From incremental automation to general task displacement

Traditional automation waves targeted specific routine tasks in manufacturing and clerical work. The current generation of AI systems differs in two structural ways. First, they are increasingly general-purpose : the same base model can be fine-tuned or prompted to perform tasks across domains, from drafting legal briefs to debugging code to producing marketing assets. Second, their performance improves with more data and compute rather than bespoke engineering, which means capability gains propagate broadly and rapidly once new model families arrive.

When observers argue that a very large share of economically valuable work in most occupations can be done by AI, they are pointing to the task composition of jobs rather than job titles. A radiologist, an accountant and a sales executive each spend much of their day on information processing: synthesising documents, interpreting signals, generating options and crafting responses. These are precisely the activities modern large language models and associated tools are increasingly competent at automating or augmenting. The same model families can then, in principle, be deployed into hundreds of millions of roles, limited primarily by integration, regulation and organisational willingness rather than by domain-specific engineering.

This is why some investors now describe the near-future workplace as one in which AI systems perform the bulk of tasks in most jobs: not eliminating all human labour, but hollowing out enough of the task bundle that one human can supervise far more output or that many roles simply no longer justify a salary. On this view, the labour market shock is not confined to a narrow band of routine manual work but extends directly into white-collar professions previously considered automation-resilient.

The timeline shock: why "faster than most believe" matters

Predictions around dates are always contestable, but the strategic issue is the gap between how quickly AI capability curves are steepening and how slowly labour-market institutions move. The claim that by around 2030 AI systems could be technically capable of doing most tasks in most jobs is not presented as a distant science-fiction scenario, but as a plausible extension of current trends in model scaling, multi-modal capabilities and robotics. Several public interventions have suggested timeframes of roughly 5 to 10 years for AI to handle about of the task content in a similarly large share of occupations, with exceptions concentrated in hands-on and complex interventional domains such as heart or brain surgery.

Whether the figure is , or is less important than the compression of the adjustment window. Education systems, retraining pathways and social insurance mechanisms typically operate on decade-scale reform cycles. Corporate IT and process change move more quickly, especially once the economic case for adoption hardens around significant cost savings and competitive pressure. The result is a structural timing mismatch: firms can move in years; governments habitually move in election cycles; individuals build careers over decades. A ten-year window in which AI becomes a general substitute for most labour tasks risks overwhelming slow-moving systems unless policy is pre-emptive rather than reactive.

Economic abundance built on labour displacement

The economic narrative accompanying these aggressive automation predictions is not one of collapse but of abundance. If AI and robotics drive the effective marginal cost of labour-intensive services towards zero, many goods and services could become dramatically cheaper. Investors making these claims argue that by the 2040s the purchasing power of a moderate income could be an order of magnitude higher than today, with housing, education, healthcare and much routine consumption available at a fraction of current cost.

The mechanism is classically deflationary. Suppose an economy has output produced by a mix of capital and labour . In a stylised production function , AI-driven capital deepening effectively shifts productive capacity so that, for many sectors, shrinks towards zero while rises, because AI systems and robots stand in for human workers. If a large portion of can be replaced by AI capital, then for the same wage bill total output can rise sharply, or for fixed output, labour requirements can collapse. Either way, the labour share of income falls while capital share increases.

The optimistic view emphasises that even if wages fall or jobs vanish, the required income to enjoy a high standard of living could fall faster. In that world, the binding constraint shifts from access to high wages to access to the abundance produced by AI. Proposals like universal basic income, sovereign AI wealth funds and near-free public services aim to recycle AI-driven returns on capital back to citizens. The challenge is less technological feasibility than political economy: who owns the AI capital, how it is taxed, and how those taxes are redistributed.

Mass underemployment as a systemic risk

The claim that mass underemployment is not a distant possibility but a likely outcome within the next decade stems from the task-level analysis of jobs combined with the economics of AI deployment. Automation decisions are rarely made with macro employment in mind; they are made at the level of firm cost structure and competitive survival. Once AI systems can reliably handle most of the value-creating tasks in a role at significantly lower cost than a human employee, boards and executives face strong incentives to restructure staffing, often aggressively.

Underemployment risks emerge even if headline unemployment remains lower. Workers may retain some work but at fewer hours, weaker bargaining power and more precarious contracts. In sectors like call centres, back-office processing, basic coding, accounting and parts of legal and medical practice, there is credible evidence that large swathes of tasks are already partially automated with today's models. As deployment scales, firms can maintain service output while reducing human-hours demanded. The macro effect is a labour market in which jobs still exist on paper but become harder to access, more fragile and less well-paid compared with the historical productivity-sharing bargain.

Several institutional analyses, from international organisations to think tanks and AI labs, now explicitly consider scenarios of double-digit unemployment or large-scale labour displacement triggered by general-purpose AI. Policy frameworks under discussion include scaled-up unemployment insurance, rapid retraining schemes, wage insurance, and various models of basic income or capital accounts that give citizens direct exposure to AI-driven equity returns. The recurring theme is that relying on traditional slow-moving welfare systems and incremental labour regulation will be inadequate if the displacement wave arrives as quickly as some technologists project.

The tax architecture problem: where will revenue come from?

One of the most concrete tensions raised by these predictions is fiscal. Today's tax systems in advanced economies are heavily reliant on labour-based revenue: income tax, payroll tax and consumption taxes funded by wage income. If AI significantly compresses the wage bill while lifting profits and capital gains, that base erodes. Yet the social demands on the state would simultaneously expand: income support, retraining, healthcare and housing assistance for those who struggle to find work in an AI-heavy economy.

Several proposals attempt to square this circle by shifting the tax base from labour to capital. One family of ideas focuses on taxing capital gains at similar or higher rates than ordinary income, especially for very high earners, on the grounds that AI-driven wealth will accrue disproportionately to holders of tech equity and intangible assets. Another emphasises a national or regional sovereign wealth fund that accumulates stakes in AI and complementary technologies, using dividends and capital appreciation to fund social transfers and public goods. A third explores explicit taxes on automated labour, sometimes framed as robot or AI usage taxes, though these raise difficult measurement and innovation-incentive questions.

The core mathematical intuition is straightforward. If aggregate labour income stagnates or falls while capital income rises sharply, and if government revenue is currently something like , maintaining or expanding public spending requires shifting the relative tax rates and . Without such adjustment, the tax base shrinks even as social demands rise. Designing that shift in a way that preserves investment incentives, avoids large-scale avoidance and remains politically legitimate is one of the defining policy design challenges of the AI era.

Work, purpose and the politics of not needing a job

Beneath the fiscal and macroeconomic arguments lies a more human question: what happens to societies built on the moral centrality of work when large numbers of people no longer need jobs to survive, or cannot find them? Some technologists argue that much modern employment is a form of economic servitude, and that freeing humans from the need to work could unleash a flowering of creativity, care and self-directed projects. Others warn that work is a core source of identity, social connection and status; strip it away without robust replacement institutions and you risk alienation, polarisation and social unrest.

Political systems are not neutral in this debate. Welfare states have historically justified support on the basis of temporary misfortune, disability or old age, not permanent structural redundancy for large swathes of the population. Expanding unconditional transfers or universal basic income raises deep questions about deservingness, free-riding and social cohesion. Meanwhile, the prospect that young children today might never need to seek traditional employment collides with educational structures still geared towards preparing people for jobs that may not exist in twenty years.

Managing this transition requires more than economic engineering. It involves rethinking education towards lifelong learning, civic participation and creative skills; reshaping urban design and community institutions to accommodate more unstructured time; and constructing new narratives of status and contribution that are not anchored solely in paid employment. The risk is a bifurcation between a small elite of AI owners and shapers and a much larger population living on transfers but with limited agency over the systems that govern their lives.

Debates, objections and empirical uncertainty

Not all economists and labour scholars accept aggressive timelines for AI-driven underemployment. Historical experience with automation shows that while specific occupations disappear, new roles emerge, and aggregate employment can remain robust or even expand. The World Economic Forum, for example, has projected large net job creation over the next decade when accounting for new roles in AI, green technologies and care, even as millions of existing jobs are displaced. International bodies such as the OECD highlight both opportunities and risks, stressing the role of policy in shaping outcomes.

Critics of the most ambitious automation forecasts raise several objections:

- Task complexity and tacit knowledge : Many jobs involve tacit, context-specific skills, emotional labour and physical presence that are harder to automate than pure information processing. Even if AI can handle of cognitive tasks, the remaining may still require humans on-site, limiting the degree of headcount reduction.

- Adoption frictions : Regulatory barriers, liability concerns, cultural resistance and integration costs can significantly slow deployment, especially in highly regulated sectors like healthcare, aviation and law.

- New demand channels : Lower costs can stimulate new demand, creating jobs in adjacent areas. Historical examples include the way automation in textile manufacturing eventually led to a much larger fashion and retail ecosystem.

- Policy dampening : Governments could choose to slow automation in critical sectors, use subsidies to encourage worker retention or mandate human involvement in key decisions.

Proponents of rapid-disruption scenarios respond that this time may be different because the technology goes after the cognitive core of professional work, is general-purpose across sectors, and scales with data and compute rather than bespoke physical investments. They also point to the increasingly software-native nature of the economy, where new products and services are digital from the outset and thus trivially automatable once models are capable. The actual path is likely to reflect elements of both views, with sectoral heterogeneity: some industries may see explosive AI-driven restructuring; others may evolve more slowly under the weight of regulation and human preference.

Why coherent policy frameworks cannot wait

Across this debate, one point of convergence is the need for structured preparation. A number of AI labs, investors and policy institutions now call for pre-emptive economic frameworks that can be activated as conditions change: enhanced labour-market statistics to monitor AI displacement in real time; scalable unemployment insurance systems; pre-authorised fiscal measures that can automatically expand support when indicators breach thresholds; and standing plans for introducing more ambitious tools such as basic income or sovereign AI funds if unemployment passes specified levels.

This contingency planning approach treats AI labour disruption similarly to other systemic risks: you do not wait for the flood to finish before designing the levees. Proposals include multi-scenario playbooks where, for instance, a unemployment scenario triggers one package of training grants and wage insurance, a scenario triggers expanded income support and sectoral transition programmes, and an unprecedented underemployment scenario opens the door to new forms of income replacement and capital redistribution. The emphasis is on building institutional muscles now, while labour markets remain mostly intact, rather than scrambling under crisis conditions.

In parallel, educational, corporate and community actors have roles to play. Workforce frameworks focused on AI readiness urge universal AI literacy, worker participation in technology deployment and flexible training pathways that can be updated as task requirements shift. Firms are encouraged to design AI adoption strategies that improve job quality and safety where possible, redeploy workers rather than simply shedding them, and share productivity gains in ways that maintain social licence. And civic discourse needs to move beyond binary narratives of utopia or dystopia towards practical questions of ownership, governance and distribution.

The underlying claim that a large portion of economically valuable work may soon be performed by machines is ultimately less a prophecy than a forcing function. It surfaces uncomfortable but unavoidable questions about how societies tax, spend, educate and define human flourishing in an age of rapidly advancing intelligence technologies. Whether or not the most aggressive timelines prove accurate, the downside risk of being unprepared for a sharp labour-market shock is large, while the upside of having a coherent framework ready is substantial. In that sense, the real wager is not about the exact percentage of jobs affected, but about whether institutions can learn to move at something closer to the speed of code.

Term: Replacement cost depreciation - Valuation

"Replacement cost depreciation is a valuation method that calculates an asset's current cost to replace, minus a deduction for physical wear and tear, age, and obsolescence. It helps determine an item's current, realistic value rather than its original purchase price." - Replacement cost depreciation - Valuation

Capital-intensive organisations constantly face a mismatch between historical prices recorded in ledgers and the economic reality of what it would now cost to recreate their operating capacity. That gap becomes most visible when assets are highly specialised, rarely traded, or subject to rapid technical change. In such settings, relying on original cost or on sparse market comparables can distort balance sheets, insurance cover and investment decisions. The practical challenge is to express a present-day, decision-useful value for an asset that may never be sold but still needs to be priced, regulated or insured as if it were replaceable.

From cost-based thinking to replacement cost depreciation

Cost-based valuation responds to this problem by treating value as anchored in what market participants would have to spend now to obtain equivalent service potential. Within that family of approaches, replacement cost depreciation focuses on the current cost of a modern equivalent asset and then deducts the loss of value from physical wear, age and different forms of obsolescence. The intent is to approximate the economic value of the remaining service capacity, rather than to reproduce historic purchase prices or pure market sale values that may not exist for specialist assets.

In practice, valuers distinguish between gross replacement cost and depreciated replacement cost. Gross replacement cost is the hypothetical cost today of constructing or acquiring a new asset that delivers the same utility as the existing one, using current prices for labour, materials, fees and profit margins. Depreciated replacement cost then applies adjustments for physical deterioration, functional shortcomings and economic disadvantages to arrive at a lower figure that better represents the existing asset in its current condition and use. In public-sector and infrastructure contexts, guidance often treats the cost approach and depreciated replacement cost as interchangeable labels for this methodology.

Substantive definition and practical meaning

Substantively, replacement cost depreciation expresses what it would cost today to replace an asset's service potential, reduced to reflect how far that potential has been consumed or impaired. The focus is not on the precise replication of every design feature but on a modern equivalent asset that can perform the same function to a similar standard. For example, a leisure centre built decades ago might be valued by asking what it would now cost to construct a contemporary facility offering comparable capacity and amenities, then deducting for the current building's age, condition and outdated layout.

This perspective matters because many specialist properties or engineering assets are seldom sold on an open market, so there is little or no transactional evidence for direct comparison. Examples include public-sector facilities, utilities, defence installations, and bespoke industrial plants. For these, replacement cost depreciation becomes a primary route for estimating fair value in financial statements, for setting regulated asset bases, or for assessing insurance sums insured.

Practically, the method is also used where the main objective is to ensure that a business could, in principle, restore its operating capacity after a loss. In insurance, replacement cost-based covers look to the cost of rebuilding or replacing with new, often with policy wordings that imply a depreciated adjustment if repairs restore but do not upgrade the asset. In regulatory and financial reporting contexts, professional standards stress that depreciated replacement cost is typically used when market-based valuation is impossible or unreliable rather than as a matter of preference.

Core steps in a replacement cost depreciation valuation

Despite variations by sector, country and standard-setter, a broadly consistent sequence is followed:

- 1. Estimate gross replacement cost. The valuer determines the current cost of constructing or acquiring a modern equivalent asset, reflecting present-day rates for materials, labour, design, fees, contingencies and profit. This can be done using unit cost rates per area or volume, or by an elemental approach that prices each component separately.

- 2. Assess land separately where relevant. For property, land is usually valued by reference to market-based comparables or residual approaches, because land generally does not depreciate in the same way as buildings. Land value is then added to the depreciated replacement cost of improvements.

- 3. Apply depreciation for physical deterioration. The asset's effective age, condition and maintenance history are used to quantify the loss of value from wear and tear, damage and material fatigue. This can include both curable defects (repairable) and incurable deterioration.

- 4. Adjust for functional obsolescence. Functional depreciation captures lost utility due to outdated design, inefficient layouts, or superseded technology that reduces output, raises costs or impairs usability despite physical soundness. For example, an industrial facility with obsolete process flows may attract a larger functional deduction than its age alone would suggest.

- 5. Adjust for economic obsolescence. Economic depreciation reflects external factors such as regulatory changes, demographic shifts, overcapacity, competition, or neighbourhood decline that reduce the economic advantage of the asset in its current use. Even a nearly new facility may suffer significant economic obsolescence if demand for its services has collapsed.

- 6. Derive depreciated replacement cost. After combining these depreciation allowances, the valuer arrives at a depreciated replacement cost figure for the improvements. For property assets, this is typically combined with the separate land value to form an overall indication of value.

Although this looks straightforward, each step embeds professional judgement about technical feasibility, market behaviour and regulatory context. Differences in assumptions about the modern equivalent design, cost baselines, or the severity of obsolescence can materially move the valuation outcome.

Mathematical specification and parameter meanings

Where a more formal representation is useful, depreciated replacement cost can be expressed in simplified form as:

where is depreciated replacement cost, is gross replacement cost, and is the total proportion of depreciation from all causes. Expanding into components gives:

subject to the constraint that . Here:

- represents physical deterioration due to age, wear, and environmental exposure.

- represents functional obsolescence from design or technological limitations.

- represents economic obsolescence from external market or regulatory conditions.

Depreciation may be modelled by reference to an asset's estimated total economic life and effective age . In a simple straight-line scheme, the proportion of physical depreciation might be approximated as:

subject to caps for condition factors or major refurbishments. More complex approaches may use diminishing-balance or survival probability models where the marginal depreciation declines or follows an empirically estimated pattern. In practice, functional and economic components are often applied as percentage deductions linked to specific identified issues rather than via a generic time-based formula.

When land is valued separately, the total value would typically be stated as:

where is market-based land value and is the depreciated replacement cost of buildings and other improvements.

Contexts of use and sectoral nuances

Professional standards for valuers emphasise that depreciated replacement cost is a method of last resort when market evidence is lacking, but it is indispensable in several domains.

- Specialised properties. Assets such as hospitals, schools, leisure centres, waste treatment plants and complex industrial facilities often have few, if any, comparable transactions in their existing use. Depreciated replacement cost provides a structured route to estimate value for financial reporting, taxation, or strategic planning.

- Public-sector asset registers. Governments and public bodies maintain large portfolios of non-market assets whose service potential needs to be quantified for accountability, performance management and fiscal analysis. Cost-based valuation methods, including depreciated replacement cost, are widely used in this context.

- Insurance and risk management. For property and equipment insurance, sums insured are typically aligned to the cost of reinstatement on a replacement basis, adjusted where appropriate for age, condition, and obsolescence. Accurately modelling depreciation ensures that cover is neither excessively generous nor dangerously inadequate.

- Regulated utilities. In some jurisdictions, regulators derive a regulated asset base using depreciated replacement cost to approximate the value of infrastructure on which allowed returns are calculated. Here, the choice of depreciation assumptions can directly influence consumer tariffs and investor returns.

Major schools of thought and methodological variants

Within the cost-based family, several currents of thought shape how replacement cost depreciation is interpreted and applied.

One important distinction is between strictly like-for-like replacement and modern equivalent replacement. Valuers increasingly favour the latter, arguing that a rational market participant would not reproduce outdated, inefficient designs where contemporary techniques deliver the same utility more cheaply or more sustainably. This modern equivalent view aligns with the idea that gross replacement cost should reflect current best practice in delivering the same service potential, not a museum-quality replica of the original asset.

Another dimension concerns how closely depreciated replacement cost is tethered to observable market prices. Some approaches emphasise internal consistency within a cost model, treating GRC and depreciation assumptions as primarily technical constructs. Others stress the need to cross-check results against any available market evidence, even if imperfect, to avoid drift from realistic trading values. Professional guidance tends to support this second view, encouraging reconciliations against any sales, income capitalisation metrics, or alternative valuation techniques where possible.

There is also debate about whether depreciated replacement cost should be interpreted as a measure of fair value in the sense used by financial reporting standards, or as a specialised notion of value in use tailored to particular stakeholders. Some argue that, provided the modern equivalent and depreciation assumptions reflect market participant perspectives, depreciated replacement cost can approximate fair value where markets are thin. Others caution that, because the method is heavily model-based and reliant on valuer judgement, it may diverge materially from the price that would actually be negotiated in a hypothetical sale, especially where the current use is not the highest and best use.

Key tensions and debates

Several recurring tensions explain why replacement cost depreciation remains a live topic in valuation discourse.

1. Objectivity versus judgement. Cost-based methods are sometimes presented as more objective than market-based valuation because they rest on observable construction costs and explicit depreciation calculations. In reality, substantial judgement enters when specifying the modern equivalent, selecting cost data, determining effective age, and quantifying functional and economic obsolescence. Different valuers can legitimately reach different answers while following the same broad guidance. This subjectivity raises concerns about comparability across entities and periods.

2. Historic cost versus replacement cost in accounting. Traditional historical cost accounting records assets at purchase price less accumulated depreciation over time. Replacement cost perspectives argue that such figures can become irrelevant for capital-intensive entities operating in environments of rapid cost inflation or technological change. Using depreciated replacement cost in financial reporting may produce balance sheets that better reflect the resources needed to maintain service capacity, but it also introduces more measurement uncertainty and potential volatility.

3. Economic substance versus legal form. In some contexts, legal or regulatory constraints limit alternative uses or disposition of an asset. The question then is whether depreciated replacement cost should assume that a buyer would pay for all the replacement costs incurred, or whether the valuation should be discounted to reflect restrictions. Practice generally requires explicit consideration of covenants, planning constraints and obligations that affect use or disposal, with appropriate adjustments to the valuation to maintain consistency with the assumed basis of value.

4. Treatment of land. Because land does not depreciate in the same way as structures, combining cost-based valuations of buildings with market-based land values can generate tensions when the land has potential for alternative, more valuable uses. Guidance therefore stresses the need to align the valuation basis with the assumed use; for example, if highest and best use would be redevelopment, a pure depreciated replacement cost of the existing improvements may overstate the economic relevance of the current configuration.

Why replacement cost depreciation still matters

Despite the growth of sophisticated income and market approaches, replacement cost depreciation continues to occupy a critical place in the valuation toolkit. For many infrastructure, public-sector and specialist industrial assets, there is simply no liquid market from which to infer value directly. Yet policymakers, regulators, investors and insurers require defensible numbers to make capital allocation, pricing and risk decisions. Depreciated replacement cost offers a principled way to bridge that gap by linking value to the economic effort required to re-establish service potential.

The method also provides a conceptual anchor in debates about economic sustainability and capital maintenance. When organisations ask how much they must reinvest to preserve operating capacity, or when regulators seek to set tariffs that allow for recovery of prudently incurred costs, replacement cost-based measures often underpin the analysis. In that sense, the approach is less about predicting transaction prices and more about maintaining the integrity of productive capital over time.

At the same time, the ongoing debates about subjectivity, fair value alignment and the treatment of obsolescence show that replacement cost depreciation cannot be applied mechanically. It demands careful articulation of assumptions, transparent documentation of methods, and critical cross-checking against whatever market or income evidence exists. Standards issued by professional bodies are increasingly focused on ensuring that users of valuations understand both the strengths and the limitations of the numbers they see.

As asset systems become more complex, digital and integrated, the challenge of valuing service potential rather than physical form will only intensify. Replacement cost depreciation, with its emphasis on modern equivalents and explicit recognition of obsolescence, provides a flexible framework for engaging with that challenge, provided its application remains disciplined, transparent and open to scrutiny.

Quote: Viktor Frankl - Author of Man's Search For Meaning

"Everything can be taken from a man but one thing: the last of the human freedoms - to choose one's attitude in any given set of circumstances, to choose one's own way." - Viktor Frankl - Author of Man's Search For Meaning

Modern life is often experienced as a negotiation with forces that appear overwhelmingly outside our control: economic shocks, political upheavals, organisational restructurings, illnesses, and random personal losses. The unsettling question beneath all of these is whether, when enough is stripped away, anything genuinely ours remains that cannot be taken or dictated by circumstance. For many philosophical traditions the answer has been yes, but the twentieth century posed this question with unprecedented brutality. It is in that harsh landscape that Viktor Frankl's claim about a final human freedom acquires its distinctive weight: it is not a theoretical flourish but a position hammered out under conditions designed to extinguish agency altogether.

The camps as a laboratory of dehumanisation

Frankl wrote on the basis of his years in Nazi concentration camps, where deliberate dehumanisation was not a side effect but a central design feature. Prisoners were transported in overcrowded cattle cars, stripped of possessions, shaved, and reduced to numbers, their prior lives and identities effaced with methodical efficiency. Hunger, exposure, arbitrary beatings, and the constant threat of selection for death created an environment of radical uncertainty. Guards and kapos enforced a regime in which moral norms were inverted: brutality was rewarded, compassion often punished.

These conditions were not merely physically destructive; they were intended to erode the inner lives of prisoners, to convert them into objects to be managed rather than subjects capable of decision. Frankl observed that many inmates succumbed to apathy and despair, a kind of inner resignation that preceded physical collapse. He described moments in which a prisoner would lie in his bunk, refusing to rise for roll call despite the risk of immediate beating or death, because an internal line had already been crossed. In such scenes the camp's strategy is laid bare: deprive individuals of any sense of meaning or effective choice until even survival ceases to matter.

Yet Frankl also reports that there were prisoners who, in the same barbed-wire world, used their meagre strength to comfort others, share a crust of bread, or refuse to inform on fellow inmates. These gestures were often materially irrational; they increased the likelihood of the giver's death. But precisely in that irrationality they demonstrated something camp logic could not account for: a capacity to place a value above survival itself. For Frankl, these acts were empirical data showing that the machinery of terror, though overwhelming, was not omnipotent. It could determine conditions, but not exhaust all human responses to those conditions.

From suffering to meaning: the architecture of logotherapy

Frankl's broader framework, logotherapy, begins from the claim that human beings are primarily driven by a "will to meaning", not by pleasure or power. Where Freud emphasised the pleasure principle and Adler a striving for superiority, Frankl argued that people are fundamentally oriented towards discovering and realising significance in their lives. When that drive is blocked, individuals experience what he called an "existential vacuum": boredom, emptiness, and a sense that life is ultimately pointless.

Logotherapy therefore focuses less on excavating the past and more on clarifying the concrete meanings available in the present and future. Frankl suggested three broad pathways: creating or accomplishing something; encountering someone or something through love, beauty, or goodness; and adopting a meaningful attitude toward unavoidable suffering. In the camps, the first two avenues were drastically restricted. Prisoners could perform labour, but it was usually pointless or exploitative. They could form bonds of friendship or love, but these were constantly threatened by separation and death.

It was the third path that remained open in nearly every circumstance: the possibility of relating to suffering in a way that preserved dignity, value, or purpose. Frankl insisted he did not romanticise pain; avoidable suffering should be removed rather than sanctified. But where suffering cannot be removed, he argued, it can still be given meaning through the stance one takes towards it. If an individual can no longer change a situation, they retain the potential to change themselves in relation to it. This is the domain in which the choice of attitude becomes decisive.

Freedom within limits: what choice of attitude does and does not claim

It is tempting to misread Frankl as denying the shaping power of environment, or as suggesting that inner attitude alone is sufficient to overcome any hardship. His own account warns against such simplifications. He acknowledged repeatedly that camp conditions exerted immense psychological pressure and that many reactions were heavily conditioned by hunger, cold, and fear. Differences in personality, prior health, random strokes of luck, and access to support networks all played roles in who survived and how.

The claim about a remaining freedom is not a denial of these determinants but an assertion that they are not total. Circumstances strongly constrain the range of possible actions and emotional reactions; they do not logically dictate a single inevitable response. Frankl counters a kind of psychological determinism that would reduce human beings to passive outputs of stimuli and drives. He insists on a wedge of autonomy: however narrow, there exists some space in which one can endorse, resist, or reframe the impulses and pressures flowing through one's life.

In therapeutic terms, this becomes a pivot from "I am nothing but my trauma, my conditioning, my diagnosis" to "these factors influence me, but they do not exhaust who I am or what I may choose to do next". Frankl did not deny unconscious motives or social forces; he rejected the idea that they fully define the person. The freedom he speaks of is not the freedom to choose any outcome, but to choose a stance in the face of outcomes one cannot fully control.

Strategic tension: agency, responsibility, and burden

There is a strategic strength and a potential hazard in framing human dignity around the ability to choose one's attitude. On the one hand, it relocates the ultimate site of agency to something that cannot be confiscated by tyrants, market crashes, or illness. This can be deeply empowering in contexts where external options are sharply limited: a political prisoner can still refuse to sign a false confession; a patient with a terminal diagnosis can still decide how to face their remaining time.

On the other hand, emphasising inner freedom risks being co-opted into narratives that downplay structural injustice. If every individual is told that their main task is simply to choose a better attitude, systemic exploitation, racism, or poverty may be reframed as mere opportunities for personal growth. Frankl's own life complicates such misuse: he was clear that the camps were an atrocity, not a neutral backdrop for character development. The inner freedom he described did not absolve perpetrators of responsibility, nor did it imply that victims had somehow failed if they succumbed to despair.

There is therefore a delicate balance between affirming agency and acknowledging limits. To deny the remaining freedom is to collapse the person into pure victimhood, which can itself be dehumanising. To absolutise that freedom is to risk blaming sufferers for not exercising it "correctly". The ethical task is to hold both truths: circumstances can be unjust and crushing, and yet within them some margin of choice persists that is morally significant but not infinite.

Debates and objections: is attitude really always available?

Philosophers and psychologists have raised several objections to the universality of Frankl's claim. One challenge concerns cases where cognitive capacity is severely impaired, such as advanced dementia, certain psychoses, or extreme brain injuries. In such situations, the ability to reflect, reframe, or commit oneself to a value-laden attitude may be diminished to the point where talk of "freedom" becomes metaphorical at best. Contemporary discussions of autonomy recognise gradations of capacity rather than a simple present-or-absent binary.

Another line of criticism draws on trauma research showing that prolonged exposure to terror can reshape neural pathways and stress responses in ways that make certain reactions almost reflexive. Individuals with post-traumatic stress disorder may experience flashbacks, dissociation, or overwhelming panic that do not feel chosen in any meaningful sense. If the nervous system is firing intensely conditioned responses, how meaningful is it to speak of selecting an attitude?

Frankl's defenders might reply that his claim is normative and existential rather than neuroscientific: he is saying that the person is not reducible to their conditioned responses and that, where even a sliver of reflective distance is available, it can be used to orient oneself differently. Moreover, the very existence of therapies that help trauma survivors gradually regain a sense of choice can be seen as partial confirmation that the capacity for inner decision is not permanently destroyed, even if it is profoundly wounded.

A further objection is political: an excessive focus on inner freedom may depoliticise suffering. If the emphasis remains on how individuals interpret oppression, attention may drift away from the institutions and power relations that produce it. Here again, Frankl's biographical and intellectual context matters. He supported efforts to rebuild humane institutions after the war and framed logotherapy as a response to widespread meaninglessness in modern mass societies, not as a replacement for social critique. The inner decision he highlighted was meant to complement, not substitute for, outer change.

Why the claim resonated: post-war existential hunger

When Man's Search for Meaning was first published, it entered a world grappling with the moral wreckage of totalitarianism and the anxieties of rapidly modernising societies. Many readers found in Frankl a voice that neither minimised horror nor surrendered to nihilism. His insistence on a remaining freedom offered a way to respond to experiences that had shattered earlier assumptions about progress and civilisation.

Unlike purely theoretical existentialist texts, Frankl's narrative interweaves abstract reflection with concrete camp episodes: a prisoner choosing to console another instead of hoarding resources; an inmate contemplating a memory of his wife as a source of inner strength; the decision to interpret a brutal joke by a guard not as the last word on human nature but as an occasion to affirm a different set of values. The book's popularity owes much to this fusion of narrative authority and philosophical argument. Readers sense that these claims were not worked out at a quiet desk but in conditions where their falsity would have been brutally exposed.

In subsequent decades, the idea that one can choose one's attitude has been absorbed into popular self-help culture, sometimes in cruder forms. Corporate workshops and motivational speeches often echo Frankl's language while detaching it from its historical and ethical context. The risk is that what was a fierce assertion of dignity against totalitarian annihilation becomes a bland injunction to "stay positive" in the face of trivial inconveniences. Recovering the depth of the original context resists this flattening.

Attitude, meaning, and contemporary crises

Despite the dangers of dilution, Frankl's insight continues to matter precisely because many contemporary crises blend external constraint with internal disorientation. Economic precarity, ecological anxiety, and rapid technological change generate forms of helplessness that are not as physically brutal as camps or gulags but can still produce an existential vacuum. People may feel that their lives are buffeted by algorithms, markets, and bureaucracies that render individual choices insignificant.

In such a landscape, the assertion that one can still choose a way of relating to circumstances re-opens the search for meaning where passive resignation threatens to take over. This does not mean adopting a forced optimism. Frankl's own later essay on "tragic optimism" argues instead for the possibility of saying yes to life despite pain, guilt, and death, precisely by treating each as an opportunity to respond with courage, responsibility, or integrity. That response may include protest, activism, or refusal, not merely internal acceptance.

For example, a person confronting serious illness might find meaning in becoming a more attentive presence for loved ones, in participating in clinical research that may benefit others, or in bearing suffering in a way that communicates something about dignity to those around them. None of these erase the injustice or randomness of the disease. They do, however, represent choices about attitude that convert a purely negative event into a context for value-realisation.

Choosing one's way: from attitude to action

It is crucial that Frankl speaks not only of choosing an attitude but of choosing "one's own way". In his framework, attitude is not a private, inward emotion isolated from behaviour. It is a stance that orients concrete decisions: whether to collaborate with cruelty or resist it, whether to give way to bitterness or to remain open to future possibilities, however limited. In logotherapy, the discovery of meaning frequently culminates in a specific task or responsibility that the individual recognises as uniquely theirs.

This emphasis guards against the critique that inner freedom is merely escapist. For Frankl, the authenticity of an attitude can be tested by the actions it generates. A chosen way might involve continuing professional work with renewed sense of service, repairing damaged relationships, or accepting sacrifices in order to remain faithful to a moral conviction. Even in severe restriction, the "way" can consist of micro-acts: a word of kindness, a refusal to demean another in order to gain advantage, a silent decision not to surrender one's inner image of a beloved person or ideal.

Here Frankl aligns with a long philosophical lineage in which freedom is not primarily the absence of constraints but the capacity to act according to values one endorses. The camps exposed a world where most external liberties had been stripped away. What remained was the possibility of orientation: to become, in Frankl's phrase, the kind of person who is worthy of their suffering, or at least not wholly determined by it.

Why it matters

Frankl's claim about a last human freedom matters because it anchors dignity in something that technological, political, or biological systems cannot fully capture or commodify. In an era where data-driven models increasingly predict and influence behaviour, there is a temptation to view persons as bundles of probabilities, their choices statistically anticipated by algorithms. Frankl's insistence on an irreducible zone of decision counters the slide from understanding patterns to accepting total predictability.

At the same time, taking his claim seriously prevents a drift into cynical fatalism. If every action is simply the result of conditioning and circumstance, moral responsibility erodes: perpetrators can blame systems, victims can be written off as inevitable casualties, and bystanders can shrug at their own inaction. Frankl's perspective restores a space in which praise, blame, admiration, and repentance remain meaningful because individuals really could have oriented themselves differently, even if doing so was costly and difficult.

Finally, the idea that attitude remains a freedom of last resort offers a resource for personal resilience that neither denies suffering nor glorifies it. It invites individuals and communities to ask, in the face of whatever cannot be changed: what is still within our power to decide? Which values will we embody here, now, under these conditions? The answer will differ from one life to another, but the very asking of those questions enacts the freedom Frankl refused to relinquish.

techfest 2026 - AI as a business model

At Techfest 2026, Marc Wilson, Managing Partner of Global Advisors, delivered a compelling opening keynote exploring the rapid evolution of artificial intelligence from a simple operational tool into a core driver of competitive strategy and business model design.

Introduced by host Tony van Niekerk, the session kicked off by framing technology not as an efficiency gimmick or a mere rollout of better chatbots, but as a fundamental decision layer capable of rewriting the rules of customer relationships, risk underwriting, and distribution across the insurance sector.

Drawing on Global Advisors’ first-hand experience as an entirely rebuilt, AI-native firm, Wilson contrasted the current AI landscape with the hype of the early 2000s .com bubble. He emphasised that while past digitisation waves focused on making existing processes faster or cheaper, today's machines insert complex inference directly into critical business steps—from automated customer agents independently renegotiating policies to upstream prevention models reshaping the very nature of risk. The address challenged industry leaders to move beyond operational testing and proactively decide where trusted intelligence should sit before market expectations are permanently altered by fast-moving competitors.

Key Takeaways:

1. The Autonomous Customer Threat: How third-party software agents are beginning to intermediate client relationships, pricing, and policy switching without your permission.
2. Three Critical Boardroom Questions: Assessing where your true value is created, who genuinely owns your customer relationship, and locating your irreplaceable strategic "moat".
3. The Reframe:Why AI is not a tool you adopt, but an economic and decision layer forming permanently underneath your corporate infrastructure.
4. The Timing Gap Risk: Why early movers who scale AI-led business model changes lock in exponential advantages that late adopters can never realistically close.
5. From Payer to Prevention Engine: Lessons from top industry performers moving from reactive claims handling to proactive, data-led risk reduction ecosystems.
6. Building the "Harness": Why 98.4% of your corporate AI focus needs to be on governance, enterprise memory, and workflow safety rather than the models themselves.

Term: EBITDA margin - Financial accounting

"EBITDA margin is a financial efficiency metric that measures a company's core operating profitability as a percentage of its total revenue, calculated by dividing EBITDA by net sales." - EBITDA margin - Financial accounting

What ultimately concerns lenders, equity investors and managers is not just how much revenue a business generates, but how efficiently that revenue is converted into operating earnings that can service debt, fund reinvestment and reward shareholders. Focusing on the profitability of the core engine of the business, stripped of financing decisions and non-cash accounting charges, is a way to separate enduring performance from noise created by capital structure, tax rules and depreciation policies.

Underlying economic issue: operating performance without capital structure noise

Companies with similar products and customers can look very different on the income statement purely because of how they are financed and how their assets are depreciated. One firm might own its factories and carry heavy depreciation; another might lease similar assets and show lower depreciation but higher rental expenses. Two otherwise comparable firms can also have very different interest expenses and tax charges because of leverage and jurisdictional tax rules. Analysts therefore need a way to compare operating profitability before the effects of financing choices and non-cash charges distort the picture.

This is where earnings before interest, taxes, depreciation and amortisation, and the associated margin, become useful. EBITDA isolates the profit generated by day-to-day operations before interest and taxes and before deducting non-cash charges such as depreciation and amortisation. Expressing that profit as a percentage of revenue provides a normalised measure of how much operating earnings are produced per unit of sales.

Substantive definition of EBITDA and EBITDA margin

EBITDA can be thought of as operating profit adjusted to remove non-cash depreciation and amortisation charges and to ignore interest and tax. In practice, there are two common construction routes, both relying on line items readily available in standard financial statements:

- Start from operating income (often labelled EBIT) and add back depreciation and amortisation: EBITDA = operating income + depreciation + amortisation.

- Start from net income and add back interest, tax, depreciation and amortisation: EBITDA = net income + interest expense + taxes + depreciation + amortisation.

In symbolic form, if is operating profit, depreciation and amortisation, one widely used specification is:

The margin then scales this earnings figure by the revenue that produced it. If is net sales or net revenue, the basic relationship is:

Analysts typically multiply this ratio by 100 to express it as a percentage. Conceptually, it answers the question: for each unit of revenue, what fraction is left as EBITDA after paying cash operating costs such as materials, labour and overhead but before interest, tax and non-cash depreciation and amortisation.

Practical calculation using financial statements

From a practical accounting perspective, the calculation is straightforward but requires careful sourcing of figures:

- Revenue or net sales: taken from the top line of the income statement, sometimes adjusted for returns, discounts and allowances to obtain net revenue.

- Operating income (EBIT): calculated as revenue minus cost of goods sold and minus operating expenses such as selling, general and administrative costs.

- Depreciation and amortisation: often disclosed either as separate line items on the income statement or in the cash flow statement under operating activities.

Using the EBIT route, the steps are:

- Compute .

- Identify total from the cash flow statement or notes.

- Calculate .

- Compute .

Alternatively, starting from net income , interest expense and tax , plus and , the construction is:

Whichever route is used, consistency over time and across peers is crucial. Analysts will usually specify explicitly which definition they adopt and ensure they apply it uniformly when comparing companies or tracking trends.

Parameter meanings and what the margin captures

The core parameters in the EBITDA margin ratio are directly linked to familiar financial statement items:

- (net sales): revenue from customers after returns and discounts, representing the scale of the firm.

- : direct costs of production or service delivery (materials, direct labour, manufacturing overhead).

- : other operating expenses such as marketing, administration and distribution.

- and : non-cash charges reflecting the allocation of past capital expenditures and acquired intangibles over their useful lives.

- : a hybrid earnings measure aiming to approximate operating cash generation before capital expenditures and financing costs.

By treating depreciation and amortisation as add-backs, the margin focuses on the cost structure of operations rather than the current accounting impact of historic investment decisions. A high EBITDA margin indicates that after paying cash operating costs, a significant portion of revenue remains available for interest, tax, reinvestment in assets and distributions to owners. A low margin suggests that operating costs absorb most of the revenue, leaving limited room for these other demands.

Relationship to other margin ratios

In margin analysis, EBITDA sits between operating and net profit margins in terms of how many costs are included. Each margin answers a subtly different question.

For capital-intensive businesses such as telecoms, airlines or manufacturers, EBITDA margins are often materially higher than operating margins because depreciation and amortisation are large. For asset-light service firms, depreciation and amortisation may be modest, so EBITDA and operating margins can be similar.

Adjustments, normalisation and non-recurring items

In serious financial analysis, EBITDA is rarely taken straight from headline figures without scrutiny. To improve comparability and focus on recurring performance, analysts often adjust EBITDA by removing non-recurring, exceptional or non-operational items. Typical adjustments include restructuring charges, one-off legal settlements, gains or losses on asset disposals and unusual impairments.

Formally, if is the unadjusted figure and represents net adjustments, adjusted EBITDA can be expressed as:

The associated margin then uses instead of the raw number. This normalised margin is used in valuation multiples and credit analysis because it is intended to capture sustainable earnings power rather than temporary distortions.

Why EBITDA margin matters for different stakeholders

For managers, the ratio provides a diagnostic of operational efficiency. Changes in EBITDA margin over time signal whether the company is improving its ability to control costs and pricing relative to revenue. Because depreciation and amortisation are excluded, the metric reacts primarily to changes in operating policies, input costs, productivity and pricing rather than shifts in accounting estimates for asset lives.

For equity investors, EBITDA margin is a building block in valuation and strategic assessment. Higher, sustainable margins often justify higher valuation multiples because they imply greater ability to generate free cash flow after necessary capital expenditure and tax. Investors also compare a companys EBITDA margin with peer averages to identify competitive advantages, cost disadvantages or signs of under-management.

For lenders and credit analysts, EBITDA is central to measures of leverage and interest coverage. Covenants and risk models frequently use ratios such as debt to EBITDA and interest coverage based on EBITDA, which implicitly rely on the margin as the link between scale and earnings. A firm with a thin EBITDA margin is more vulnerable to revenue shortfalls because a small decline in sales can quickly erode the earnings base needed to service debt.

Industry patterns and benchmarks

What counts as a strong or weak EBITDA margin depends heavily on industry structure, capital intensity and competitive dynamics. Some sources suggest that a margin above 10 % is broadly viewed as positive, but this is highly context-specific. Consumer packaged goods, software and high-value business services often achieve margins significantly higher than this, while retail, airlines or commodity producers may operate at structurally lower margins due to intense competition and high variable costs.

Because of this variation, analysts compare a companys EBITDA margin chiefly against:

- Its own history, to identify improvement or deterioration in operating efficiency.

- Direct peers in the same industry and geographical region.

- Implicit targets embedded in management guidance or strategic plans.

A sudden divergence from peers may signal strategic missteps, cost problems or, conversely, successful differentiation and operational excellence.

Limitations and critiques

Despite its popularity, EBITDA margin is not a complete measure of profitability or cash generation, and over-reliance on it can be misleading. Critics emphasise several weaknesses:

- Ignores capital intensity: By adding back depreciation and amortisation, the metric downplays the fact that many businesses must continually reinvest large sums to maintain their asset base. Two firms with similar EBITDA margins can have very different free cash flow profiles if one requires heavy ongoing capital expenditure.

- Excludes working capital needs: EBITDA does not account for changes in inventories, receivables or payables. Businesses that must finance large working capital swings to support growth can have healthy EBITDA margins but strained cash flows.

- Can be manipulated through adjustments: Aggressive use of adjusted EBITDA, with numerous add-backs, can inflate the margin and obscure genuine operating weakness. Investors need to scrutinise the nature and persistence of adjustments.

- Omits interest and tax realities: For highly leveraged firms, ignoring interest expense can create a false sense of security. A strong EBITDA margin does not guarantee that the firm can meet its cash interest obligations once debt service is considered.

This is why sophisticated analysis typically pairs EBITDA margin with other metrics such as operating margin, net margin, free cash flow and return on invested capital, as well as detailed assessment of capital expenditure and working capital requirements.

EBITDA margin in valuation and transaction analysis

In corporate finance and deal-making, EBITDA margin underpins many practical tools. Valuation multiples, particularly enterprise value to EBITDA, implicitly treat EBITDA as a proxy for operating cash flow available to all investors (debt and equity). Because enterprise value is independent of capital structure, matching it to EBITDA yields a capital-structure-neutral valuation metric. A higher EBITDA margin generally supports higher enterprise value multiples, all else equal, especially when margins are durable and supported by competitive advantages.

In mergers and acquisitions, potential cost synergies are often modelled as incremental improvements in EBITDA margin post-transaction. For example, if a combined entity can reduce overlapping overhead or secure better procurement terms, the forecast case may assume that EBITDA as a proportion of revenue rises by several percentage points. Small percentage improvements can translate into large value creation when applied to substantial revenue bases.

Schools of thought: enthusiasm versus scepticism

There is a long-running debate between practitioners who view EBITDA margin as a central indicator of operating success and those who argue it can obscure economic reality.

Supporters stress that the margin cleans away distortions created by differences in capital structure, tax regimes and accounting policies for depreciation and amortisation. They argue it allows an analyst to concentrate on the efficiency of the operating model itself: how much of each unit of sales remains after paying suppliers and staff. Especially in cross-border or cross-industry comparisons where tax and accounting conventions vary, this standardisation is attractive.

Critics counter that by excluding depreciation and amortisation, the metric risks implying that assets do not wear out or that intangible investments are costless. For asset-heavy industries, they argue, replacing plant and equipment absorbs such a large portion of cash that viewing profitability before these costs is of limited relevance. Some also question the widespread use of adjusted EBITDA margin, where repeated classification of material costs as "non-recurring" can systematically overstate underlying performance.

Between these poles lies a more nuanced school of thought: use EBITDA margin, but only alongside robust analysis of capital intensity, cash flows and economic returns. In that view, the metric is a useful lens on operating efficiency, not a complete picture of value creation.

Why the concept continues to matter

Despite its limitations, EBITDA margin remains deeply embedded in financial markets, corporate reporting and internal performance management. It persists partly because it bridges the language of accounting and the economics of cash generation. Managers can influence it directly through pricing, cost control and mix decisions, while investors can use it to compare companies on a broadly consistent basis even when their financing choices or tax environments differ.

The concept has also adapted. Many firms now disclose segment-level EBITDA margins, allowing stakeholders to see which business lines generate the strongest operating earnings relative to revenue. Credit agreements use covenant tests based on EBITDA levels and margins, shaping behaviour by constraining leverage when profitability deteriorates. Private equity practitioners build value-creation plans around raising portfolio companies EBITDA margins via efficiency improvements and strategic repositioning.

In contemporary analysis, the most effective use of EBITDA margin is as one element within a multi-metric framework. It highlights the proportion of revenue that survives cash operating costs, providing insight into cost structure and operational leverage. Combined with measures of capital expenditure, working capital needs and cost of capital, it helps form a comprehensive view of whether a business model genuinely creates value over time, rather than merely reporting high accounting profits in a single period.