Select Page

Artificial Intelligence

Lessons from our journey towards AI-native

AI is becoming a new economic layer, making machine intelligence cheaper, more capable and more widely available. The strategic opportunity lies beyond efficiency: in how organisations redesign business models, combine models with data, systems and people, and prepare for second-order effects that could reshape entire industries. As AI raises the floor, competitive advantage will increasingly depend on architecture, talent, judgement, enterprise memory and values.

How is AI Amplifying you?

About two and a half years ago, we made a decision at Global Advisors that, looking back, was probably more consequential than we understood at the time.

We were approaching our twentieth anniversary. We had a consulting firm that worked: good clients, experienced people, methods developed over many years and a substantial body of past work. Then generative AI arrived, and I became increasingly convinced that strategy consulting was going to change more over the next few years than it had over the previous few decades.

So we went all in. That meant taking people, money and management attention away from a business that was already working. We turned away some work, slowed recruitment, bought hardware that started ageing while we were still learning what to do with it, moved our knowledge environment, deployed our own models and rebuilt parts of the technology stack more than once. At times I wondered whether we had moved too early. Then another model would arrive, or a capability would improve much faster than expected, and I would worry that we were already late.

I suspect many chief executives recognise that tension. You cannot suspend a functioning organisation while you investigate a technology moving from month to month. And yet the technology and market do not wait for you to become comfortable.

One idea has become increasingly useful to us through this journey: AI is an amplifier. It increases the product of our effort and gives people access to more knowledge, analytical capacity and ways to create. It also gives greater reach to poor information, badly designed processes and weak judgement. The interesting part of the amplifier is what happens to the person or organisation behind it.

LESSONS FROM OUR JOURNEY TOWARDS AI-NATIVE

An advert about being human

A few weeks ago we tried to express that in an advert. We deliberately made an advert about AI that mostly showed what it meant to be human. A young woman stands on a sidewalk. The camera moves into her eye and through memories from her life: being a baby, school, varsity, graduation, relationships and loss. AI only appears at the end.

Afterwards, one of South Africa’s top recruiters sent me a message. His point was that the video walks through a whole life before mentioning AI because AI amplifies what is already there; it does not build the person. He connected that to hiring. Now that everyone’s CV and cover letter can be polished by AI, the polish tells you much less. What stands out is the person who knows their own story and can back it up when challenged. His conclusion was that AI raises the floor for everyone, which makes the human substance underneath it more important.

My response was: “Bingo.”

LESSONS FROM OUR JOURNEY TOWARDS AI-NATIVE

From AI to the inference economy

Once a capability becomes abundant, its value as a differentiator falls. That gets us surprisingly close to the economics of AI.

I think AI belongs alongside a small number of developments that changed the economic environment in which organisations operated: the railways, industrialisation, electricity and the internet. AI acts directly on language, knowledge, analysis, images, software and decisions. Consider how much of a modern organisation consists of people reading, interpreting, comparing, writing, judging and communicating. AI can participate in an enormous amount of that activity.
 
We have started referring to this at Global Advisors as the inference economy.
I think AI belongs alongside a small number of developments that changed the economic environment in which organisations operated: the railways, industrialisation, electricity and the internet. AI acts directly on language, knowledge, analysis, images, software and decisions. Consider how much of a modern organisation consists of people reading, interpreting, comparing, writing, judging and communicating. AI can participate in an enormous amount of that activity. We have started referring to this at Global Advisors as the inference economy.

Machine inference has become something that can be purchased on demand. A machine can interpret a document, recognise a pattern, write code, compare alternatives or propose a course of action, and the cost of those cognitive acts is falling very quickly.

For most of corporate history, cognitive effort has been expensive because it has been attached to human time. Once a meaningful proportion of that effort can be purchased cheaply and at enormous scale, the economics begin to change.
 
The models are also showing that they are more than cheaper substitutes for work people already do. AlphaFold changed the scale and speed at which protein structure prediction could be approached. We are seeing similarly surprising work in mathematics and scientific reasoning. The more interesting possibility is that AI lets us solve different problems, or familiar problems in ways that were previously impractical.
 
That is why efficiency cannot be the end of the strategy. Electricity created its larger gains when organisations redesigned factories and products around electrical power rather than simply replacing a steam engine with an electric motor. We may make the same mistake with AI if we focus too narrowly on inserting a model into each existing task.

LESSONS FROM OUR JOURNEY TOWARDS AI-NATIVE

Productivity raises the floor, but unevenly

The immediate gains are real.

At Global Advisors, consultants and developers can move from a blank page or idea to a credible starting point much faster, and small teams can examine amounts of material that previously required far more people.

But competitors will have access to the same foundation models. Technology providers will put these capabilities into standard software. Employees and customers will have them as well. If every company in an industry reduces the cost of a service, competition eventually moves some of that value towards customers through lower prices, faster turnaround and higher expectations. The productivity benefit becomes part of the standard required to compete.

The uplift is also uneven. AI capability itself has a jagged edge: it can perform extraordinarily well on one task and fail on another that looks deceptively similar.

People have a jagged edge too. Matthew Prince at Cloudflare has described some employees becoming multiples more productive with AI while others get much less leverage from the same broad class of tools.

That makes deployment and management harder. Giving everyone a licence does not produce a common productivity gain. People differ in how well they frame problems, decompose work, choose tools, build repeatable workflows, validate outputs and recognise when the model has crossed the boundary of its capability.

This also makes talent more important. If AI is an amplifier, the return on attracting exceptional people can increase because their judgement, curiosity, domain knowledge and agency now have much greater reach. Firms need to attract and retain people who know how to work at that frontier, and create incentives for their individual breakthroughs to become organisational capability rather than private advantage.

LESSONS FROM OUR JOURNEY TOWARDS AI-NATIVE

Where some of the magic lies

The model is only one component of an AI-native organisation. Some of the most interesting value comes from how everything is put together: data, deterministic systems, workflows, models, retrieval, enterprise memory, ontologies, model harnesses and people across the organisation, suppliers and customers.

At Global Advisors we have gradually built an architecture around those components. We are still learning how they should fit together, and so is everyone else.

Reference architectures will emerge and vendors will package more of the complexity, but I think there will be a significant period in which AI-native ingenuity and agility count for a great deal.

The organisation that understands its problem deeply enough to combine these components in an unusual and effective way can build something a competitor cannot reproduce simply by buying the same model.

Think about underwriting. A superficial implementation might ask a general-purpose model to summarise an application. A more developed architecture could combine customer and medical information, historical claims, deterministic underwriting rules, actuarial models, an ontology of conditions and risks, inference over unstructured documents and an experienced underwriter who retains responsibility for the consequential decision. Feedback from that decision then flows back into the system.

 

Think about underwriting. A superficial implementation might ask a general-purpose model to summarise an application. A more developed architecture could combine customer and medical information, historical claims, deterministic underwriting rules, actuarial models, an ontology of conditions and risks, inference over unstructured documents and an experienced underwriter who retains responsibility for the consequential decision. Feedback from that decision then flows back into the system. Now you are creating a learning system. The model may be available to everyone; the architecture, context and accumulated feedback are specific to the insurer.

Now you are creating a learning system. The model may be available to everyone; the architecture, context and accumulated feedback are specific to the insurer.

The same applies to us. We can buy access to frontier models, but our opportunity comes from connecting them with over twenty years of work, our problem-solving methods, quantitative tools, research processes, workflow, client experience and the judgement of experienced consultants. The system becomes an expression of the firm.

LESSONS FROM OUR JOURNEY TOWARDS AI-NATIVE

Second-order effects are where strategy earns its place

Most AI programmes begin with the current business: where can we save cost, remove delay or automate repetitive work? The effects that matter most may occur several steps away.
 
AlphaFold is useful again. Demis Hassabis has spoken about AI accelerating drug discovery, contributing to cures for disease and potentially extending healthy life. Nobody knows how quickly that happens, but strategy cannot wait for certainty before considering the consequences.
 
If longevity improves significantly faster than expected, a life insurer could see lower claims on some life products while annuity and pension liabilities rise. Existing books and reinsurance assumptions could change. Underwriting could become a continuous health relationship rather than a decision made mainly at inception. A medical aid could see a cure remove years of recurring treatment cost while introducing a very large upfront expense.
 
Those effects could dwarf the benefit of automating claims administration.
 
The same logic applies elsewhere. Materials breakthroughs could affect mining, manufacturing and energy. AI agents acting for customers could compare products, read contracts and negotiate continuously, reducing information asymmetry and shifting power across industries. A company can implement AI extremely well inside its existing processes and still find that the more important change happened outside its organisational boundary.

LESSONS FROM OUR JOURNEY TOWARDS AI-NATIVE

Strategy under uncertainty, including a crash

We do not know which developments will occur, how quickly they will arrive or how they will interact. That is where scenario planning, portfolios and real options become more useful than a single confident forecast.
 
One scenario that belongs in the set is a major AI market correction. I have no idea whether it happens, when it happens or how severe it might be. But there is enough capital, expectation and market value concentrated around AI that ignoring the possibility would be poor risk management.
 
A crash would not simply mean lower technology share prices. Suppliers could fail, infrastructure could reprice and capital for new capacity and frontier-model development could become scarce. Some compute or talent might become cheaper while other capabilities become more concentrated or less available. The wider effects on investment, wealth, customers and suppliers could be significant.
 
I do not think AI disappears in that scenario. The dot-com crash did not make the internet disappear; it changed ownership, capital availability, infrastructure economics and the path of deployment. An AI crash could do something similar.
 
Leadership teams therefore need to ask whether their data and workflows are portable, where they depend on a single supplier, which investments rely on continued cost declines and which opportunities might become more attractive after a repricing.
 
This is where Quantified Strategy matters to us. We can work with ranges, attach probabilities to scenarios, compare expected values and design real options. The numbers do not remove uncertainty; they force us to make the assumptions visible.

 

LESSONS FROM OUR JOURNEY TOWARDS AI-NATIVE

Memory, people and the human premium

If intelligence becomes widely available through similar models, proprietary context becomes more interesting. The more valuable asset may be enterprise memory: institutional experience, past decisions, customer context, methods, outcomes and feedback organised so that both people and machines can use them.
 
Global Advisors has over twenty years of work behind it. AI gives us the possibility of making that experience more accessible to a consultant working today, while every new assignment adds to the memory available to the next team. Knowledge can begin to compound.
 
Your competitor can buy the model. They cannot buy twenty years of your relationships, customer experience, operating lessons and judgement. But you only gain an advantage if you organise those things well enough to use them.
 
The same economic logic makes human qualities more valuable. If machine inference becomes cheaper and more abundant, scarce complements such as problem framing, domain expertise, judgement, trust, courage and ethical responsibility become more important.
 
AI can produce a competent analysis quickly. Someone still needs to know which analysis is worth doing and whether the answer makes sense. A model can present a chief executive with options, but it does not carry professional or moral accountability for the decision. And when history stops being a reliable guide because of a crisis, scientific breakthrough, market crash or genuinely new business model, people need to recognise that the pattern has broken.
 
Talent strategy therefore becomes part of AI strategy. Recruitment has to get beneath AI-polished applications. Development has to teach people to work across a moving jagged frontier. Retention has to recognise that the best AI-amplified people may create disproportionate value, including value that appears through the systems and colleagues they improve.

LESSONS FROM OUR JOURNEY TOWARDS AI-NATIVE

What Global Advisors has chosen to do

Our own experience has clarified where we want to play. We do not intend to become another large systems implementer. There are firms better placed to migrate data estates, implement platforms and run complex integrations, and we will partner with them where clients need that capability.
 
We remain focused on the consequential decisions: how AI changes industry economics; which second-order effects matter; where architecture can create advantage; how customers, suppliers and agents shift ecosystem power; where enterprise memory can compound; which decisions should use inference, deterministic control or accountable human judgement; and which futures the board should prepare for.
 
Then we quantify the choices. We build scenarios, attach probabilities, examine expected values, construct portfolios and real options, and sequence capital rather than treating AI as one enormous programme.
 
Our work can extend into incubation while new capability and business models are tested, and into acceleration once a strategic choice needs to become real through organisation, measures, governance and execution. We see both as ways of making strategy happen.
 
So when I sit with chief executives today, I spend less time asking how many AI use cases they have and more time examining the assumptions beneath the business.
 
What happens when an expensive cognitive task becomes nearly free? What happens when customers have roughly the same intelligence as employees? What happens when an agent chooses suppliers on the customer’s behalf? What happens when science changes mortality, materials or energy economics? What happens if the AI investment cycle crashes? Which parts of the business are protected by genuine capability, and which mainly by friction that AI may remove?
 
And then there is the human question: what are you amplifying?
 
If you have strong people, valuable institutional knowledge, good judgement, clear values and a business model worth building, AI gives you the possibility of extending those things far beyond their previous limits.
 
Once everybody has access to the polish, we look for the substance underneath. I think recruits, customers, employees and investors will increasingly do the same with organisations.
 
Models will become more capable. Inference will become cheaper. Reference architectures will mature. The harder things to copy will be how you put those components together and what you bring to them: your knowledge, relationships, architecture, people, agility, judgement, values and ethics.
 
Those are the things worth amplifying.
 
How is AI amplifying you?

Get In Touch

16th Floor, The Forum, 2 Maude Street, Sandton, Johannesburg, South Africa
+27114616371

Global Advisors | Quantified Strategy Consulting
error: Content is protected !!