“If you don’t have an AI strategy, you’re going to die in the world that’s coming.” – Devin Wenig – Former CEO, eBay

Competitive survival increasingly depends on whether a company can turn artificial intelligence from an isolated experiment into an operating capability. The danger is not simply that a rival may adopt a clever chatbot first. It is that competitors may redesign pricing, customer service, product discovery, fraud prevention, software development and decision-making while slower organisations continue to treat AI as a discretionary technology project. The resulting gap can widen through better data, faster feedback and lower operating costs, making recovery more difficult over time.

The strategic force behind this argument lies in the difference between using AI and building an AI strategy. A tool can improve an individual task, such as drafting copy or summarising research. A strategy connects technology to a defensible business objective, identifies the data and workflows required, assigns responsibility, and establishes measures for value and risk. Without those connections, companies accumulate demonstrations rather than capabilities. Employees may use consumer tools informally, departments may purchase incompatible systems, and executives may celebrate activity without knowing whether performance has improved.

Devin Wenig made the remark in the context of a business environment shaped by platform economics and rapid technological change. As the former chief executive of eBay, he operated in a sector where search, recommendations, trust, payments, logistics and personalisation directly affect transactions. These functions are especially suitable for machine learning because every interaction can generate signals about intent, relevance and risk. The broader lesson is not that every organisation must become an AI laboratory. It is that firms whose customer experience or cost base can be materially improved by software cannot assume that existing advantages will remain stable.

Evidence of widespread adoption supports the urgency, but it also complicates the dramatic language. McKinsey’s 2025 global survey reported that 88 percent of respondents said their organisations regularly used AI in at least one business function, while nearly two-thirds had not yet begun scaling AI across the enterprise.³ The implication is a market divided between experimentation and integration. Many firms have access to similar models, yet far fewer have redesigned work around them. Access alone therefore offers limited protection; execution, proprietary data, process knowledge and organisational discipline determine whether adoption becomes an advantage.

Workflow redesign is the crucial middle layer. If a company simply adds an AI assistant to an unchanged process, it may save small amounts of employee time while preserving approval bottlenecks, duplicated data entry and poor incentives. More substantial gains arise when responsibilities are reconsidered from end to end. A support team might combine automated triage with human escalation, a retailer might connect demand forecasts to inventory decisions, and a software organisation might integrate code generation with testing and security review. The objective is not to remove people indiscriminately, but to allocate human judgement where it has the greatest economic and ethical value.

Why urgency can produce bad strategy

Survival rhetoric can motivate boards to act, but it can also encourage waste. Fear of falling behind may lead executives to fund fashionable pilots without a clear customer problem, buy overlapping platforms, or deploy systems before data quality and accountability are ready. The technology may then generate inaccurate recommendations, expose confidential information or create inconsistent treatment of customers. NIST’s AI Risk Management Framework recommends a structured approach built around governing, mapping, measuring and managing risks, with a generative AI profile addressing issues such as fabricated outputs, harmful content, privacy concerns and security threats.^1,2 That guidance makes clear that speed and control are not opposing goals; reliable controls are prerequisites for sustainable scale.

There is also a legitimate objection to the claim’s implied certainty. AI will not eliminate the need for industry knowledge, capital allocation, strong brands or sound management. Some applications have weak economics because inference costs, integration work and human review outweigh the benefit. Other sectors face legal, safety or cultural constraints that make automation inappropriate. McKinsey’s findings show this tension: although AI use is broad, only 39 percent of respondents reported an enterprise-level EBIT impact, suggesting that adoption has advanced faster than measurable financial returns.³ A company can therefore have an AI strategy and still fail if it chooses low-value use cases or mistakes usage statistics for productivity.

A credible strategy begins with a portfolio rather than a slogan. Leaders should identify a small number of high-value workflows, rank them by customer impact and feasibility, and define a baseline before deployment. The baseline might involve resolution time, conversion, defect rates, fraud losses or employee capacity. Results should then be tested against that baseline, with attention to quality and unintended consequences as well as cost. Projects that improve a meaningful metric can receive further investment; projects that cannot demonstrate value should be stopped, redesigned or kept at experimental scale.

Data governance becomes more important as systems move from assistance to action. Models depend on the relevance, legality and integrity of the information supplied to them, while automated decisions can amplify historical bias or create new vulnerabilities. Organisations need clear rules for access, retention, provenance, evaluation and incident response. They also need named owners who can answer what a system is permitted to do, when a person must intervene and how a customer can challenge an outcome. NIST describes trustworthy systems in terms including validity, safety, security, accountability, transparency, explainability, privacy and fairness.^1,9 These are not merely compliance categories; failures in any of them can damage revenue and trust.

The competitive question is consequently less whether a company has purchased AI than whether it can learn faster with it. Organisations that build reusable data foundations, measured workflows and responsible deployment routines may compound their advantage as each implementation produces better operational knowledge. Those that chase isolated demonstrations may discover that their rivals have changed the basis of competition before the technology appears in conventional performance reports. The warning, read carefully, is not a prediction that every non-adopter will literally disappear. It is a challenge to treat AI as a matter of strategic design, organisational learning and accountability, because the firms that postpone those decisions may eventually find that the market has made them on their behalf.

 

References

1. AI Risk Management Framework – 2021-07-12 – https://www.nist.gov/itl/ai-risk-management-framework

2. Artificial Intelligence Risk Management Framework – 2024-07-26 – https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence

3. The state of AI in 2025: Agents, innovation, and transformation – 2025-11-05 – https://www.mckinsey.com/capabilities/growth-marketing-and-sales/our-insights/the-state-of-ai-2025

4. AI RMF – AIRC – 2026-06-18 – https://airc.nist.gov/airmf-resources/airmf/

5. nvlpubs.nist.gov · nistpubs · aiArtificial Intelligence Risk Management Framework: Generative …… – https://nvlpubs.nist.gov/nistpubs/ai/NIST.AI.600-1.pdf

6. [PDF] The state of AI – McKinsey – https://www.mckinsey.com/~/media/mckinsey/business%20functions/quantumblack/our%20insights/the%20state%20of%20ai/2025/the-state-of-ai-how-organizations-are-rewiring-to-capture-value_final.pdf

7. Resources – 2021-07-13 – https://www.nist.gov/itl/ai-risk-management-framework/ai-risk-management-framework-resources

8. Technical Reports – AIRC – NIST AI Resource Center – 2019-08-09 – https://airc.nist.gov/technical-reports/

9. How to apply the NIST risk framework to GenAI – Grant Thornton – 2024-06-27 – https://www.grantthornton.com/insights/articles/advisory/2024/how-to-apply-the-nist-risk-framework-to-genai

10. Department of Commerce Announces New Guidance … – 2024-07-26 – https://www.nist.gov/news-events/news/2024/07/department-commerce-announces-new-guidance-tools-270-days-following