Insights / Enterprise strategy / 31 October 2026

When AI Changes the World Around the Enterprise

Making business decisions when AI is reshaping both what enterprises can do and the conditions in which they operate.

By Abhishek Sinha · Originally published on the author's personal site · 9 min read

A sound enterprise decision can become unsuitable without having been unsound when it was made. Technology leaders know this from outsourcing agreements negotiated around particular cost assumptions, operating models designed for an established market, or platforms selected when the alternatives looked very different. Circumstances moved, while commitments made for good reasons remained in place.

Today, another procession of reports and frameworks advises enterprises on how to respond to AI. Redesign workflows, rethink talent, reorganise around agents, move beyond productivity. Much of this advice is thoughtful and grounded in research. Nor is the procession unfamiliar. Virtualisation, cloud computing, automation and digital transformation all came with compelling prescriptions, and companies pursued them at different speeds and for different purposes. The variation was not simply a measure of foresight or resistance. Each technology encountered an organisation with its own obligations, capabilities and priorities.

AI encounters those realities too, but it is also influencing the world beyond the enterprise. Its capabilities are changing alongside competitive possibilities, customer attitudes, employment expectations and regulatory choices. Enterprises must therefore assess not only what AI makes feasible, but whether an opportunity will remain attractive in conditions that are themselves evolving.

The enterprise is not a blank slate

Legacy systems, entrenched practices and organisational silos are frequently offered as explanations when technology programmes disappoint. They can certainly obstruct progress, but they are an incomplete explanation of why businesses change differently. Existing arrangements often reflect years of experience, customer obligations, regulatory requirements and hard-won expertise. Some have outlived their purpose; others remain because they work, even when their value is not obvious from an organisational chart or process map.

I have seen enterprise project-management systems give senior leaders useful financial and operational visibility while project teams continued to manage their daily work in spreadsheets and specialist tools. The central system served a legitimate purpose, but the work required to keep it current could fall on people who received relatively little direct benefit. Describing their behaviour as resistance would miss the difference between how management intended the system to work and how its users found it practical to operate.

AI initiatives face similar questions about whose work improves and where value actually appears. Faster proposal preparation does not necessarily raise win rates. Lower call-handling costs need not improve customer relationships. For a small business, however, affordable research, software development or specialist analysis may make an entirely new service possible. A larger organisation with existing teams and systems may see the same capability primarily as an opportunity to improve cost, quality or capacity.

These differences extend to management priorities. A growing company looking for new markets, a mature business protecting margins and a regulated institution preserving operational trust can reasonably approach the same opportunity differently. None can judge a proposal solely by what the technology does in isolation. Equally, organisational history offers no automatic defence of the status quo. Established practices can protect valuable knowledge or conceal outdated assumptions, and caution can become costly if a market moves before the organisation is ready. The pace of change tells us little about the quality of a decision without its context.

When the environment changes with the technology

Enterprises have always faced uncertainty. The internet redrew competitive boundaries; globalisation and outsourcing changed the economics of work; financial crises and regulation disrupted carefully prepared plans. The experience accumulated through those changes remains useful. What is harder to assess today is how several developments associated with AI will interact over time, including developments outside the control of any individual enterprise.

The technology is still changing. New models can alter the cost, reliability and feasibility of activities that looked impractical only months earlier, although their performance remains uneven across tasks. Employees find uses outside formal project plans, while AI systems are increasingly used to interpret information, shape recommendations and sometimes act across applications. An investment case developed around one level of capability may need examination when that level changes.

The competitive consequences may be more far-reaching than improvements to existing tasks. In a 2026 CEPR working paper, Jan Bena, Bo Bian and Mariassunta Giannetti report approximately 20% more startup formation in industries with greater exposure to generative AI than in less-exposed industries. The new firms tend to be smaller, and the finding does not tell us how many will become enduring competitors. It does, however, suggest a change in who can enter markets and which businesses become feasible. A customer may find that work once outsourced to a specialist provider is now within reach; a small entrant may offer a service that previously required a large organisation. A company focused on making its existing operations cheaper could find that the more consequential change is taking place in the market it serves.

Customer and employee responses introduce another dimension. I recently spoke with someone in the UK who had cancelled an economically attractive food-delivery subscription because she disliked what she perceived as the company's growing use of AI, including in its advertising. One person's reaction says nothing about the prevalence of that view, but it shows how the perceived means of providing a service can matter to a customer, even when its price and convenience remain appealing. Other customers may welcome AI-generated services, better personalisation and lower costs.

Public attitudes are mixed rather than uniformly supportive or hostile. Stanford's 2026 AI Index reports that the global share of respondents who saw more benefits than drawbacks from AI products and services rose from 55% in 2024 to 59% in 2025, while 52% said such products made them nervous. People can appreciate AI's usefulness and still be concerned about employment, privacy, creativity or authenticity. Students considering careers and employees contemplating changing roles are making decisions amid those concerns. Job applicants may question whether automated screening gives them a fair hearing, sometimes without knowing how a recruitment decision was made. Such perceptions can shape behaviour and trust before the eventual effects of the technology are clear.

Policy is moving as well, and not always towards tighter restrictions. Under the EU AI Act, qualifying systems used in recruitment and employment are classified as high-risk. In July 2026, an enacted amendment moved the application date for relevant Annex III obligations from August 2026 to December 2027. A business planning an HR system around the earlier timetable would have had to revise its compliance schedule, even though other responsibilities did not disappear. The change illustrates that regulation responds to the practical difficulties of governing AI as well as to concerns about its effects. Enterprises operating across jurisdictions must interpret both existing duties and the direction in which policy is moving.

The relationship between enterprise decisions and their environment also runs in the other direction. Widespread use of AI-generated advertising can influence what customers consider authentic. Automated recruitment can change how people prepare applications and how employers evaluate them, prompting further scrutiny of fairness. AI customer-service initiatives can alter expectations of speed, price and access to a human. One firm's decision may provoke competitive imitation, public criticism, new offerings or calls for regulatory intervention, changing the context for the next decision made by that firm and its peers.

These feedback effects need not be adverse. Greater familiarity may strengthen acceptance, and an unexpected use may create a market that neither customers nor businesses previously imagined. But a technically successful AI deployment can become less attractive if customers object to its use, just as a modest pilot can become strategically important if market conditions change. Research is accumulating on AI's technical, economic and social effects. We have much less long-term evidence about how those effects will combine across industries, how quickly they will settle and which assumptions made today will survive.

Making commitments while assumptions evolve

Businesses cannot postpone decisions until that evidence arrives. They must invest, compete, employ people and keep essential operations running. Nor can they redesign their plans whenever a new model is announced or a survey reports a change in public sentiment. Continuity matters, particularly when reliable services depend on complex arrangements built over years. The question is which changes in circumstances are material enough to revisit a decision.

Long-term outsourcing and technology programmes have taught me something about the limits of even careful planning. Agreements often provide for shifts in volumes, pricing, scope and service obligations because neither party expects conditions to stay fixed. Those provisions help, but they cannot anticipate every change in the economic rationale for an arrangement. AI may make an activity more economical in-house, alter the skills required of a supplier, or enable a completely different service. Meanwhile, a customer expectation or regulatory requirement may change even if the technology performs exactly as specified.

There is a meaningful distinction between a reversible experiment and a commitment that reshapes staffing, contracts or foundational systems. For the latter, an enterprise can be explicit about the assumptions supporting its choice, including those about competitors, customer behaviour and regulation, and about the evidence that would justify looking again. A decision not to proceed may deserve the same discipline: a proposal rejected for poor reliability or excessive cost can become sensible if either condition changes.

That evidence will not always arrive through formal strategy reviews. It may surface in customer conversations, complaints, operational workarounds, supplier negotiations or questions raised by employees and regulators. Making use of it calls for judgement about relevance and credibility rather than simply collecting more information. AI may help connect observations from across the business, but its summaries and recommendations must themselves be examined with care. Excessive deliberation has a cost too: an organisation can miss an opportunity while waiting for certainty that will never be available.

The choices ahead

Organisational learning, adaptation and the questioning of established assumptions have been studied for decades. This is not an argument that those disciplines have become obsolete or that AI requires an entirely new management theory. Their importance is easier to recognise when the environment changes in several connected ways and when the enterprise's own actions may influence what happens next.

The possibilities are substantial. AI can make investigation, creation and coordination affordable to more businesses, permit new services and offer established enterprises options that would once have been difficult to justify. Some will benefit most from improving what they already do; others may find reasons to change direction. Still others will judge that a promising capability does not yet warrant the disruption it would bring. Each choice has to be understood against the organisation's circumstances and the wider conditions in which it expects to operate.

The questions worth examining extend beyond the next technology roadmap. What might a business discover about its customers, competitors or own ways of working as AI becomes commonplace? Which signals would challenge the assumptions behind a major investment, and which deserve to be treated as passing noise? How should leaders balance the confidence required to make commitments with the curiosity to recognise when circumstances have changed? These questions do not have one answer for every enterprise, and the evidence is still developing. They seem worth pursuing alongside the increasingly sophisticated advice about what organisations should do next.

Questions arising from the evidence

Can AI change an enterprise that has not adopted it?

Yes. AI may alter the economics of competing services, enable new entrants, change what customers expect and influence employee behaviour or regulation. Those developments can affect a business’s prospects even if it deploys no AI systems of its own. Their scale and direction remain uncertain and vary with the sector and circumstances.

Does slower AI adoption necessarily signal resistance or poor judgement?

No. Established systems and practices may embody legitimate obligations, expertise and investments, and organisations can reasonably assess the same opportunity differently. Equally, the fact that an arrangement once made sense does not guarantee that its assumptions remain sound. The pace of adoption alone tells us little about the quality of the underlying decision.

When is an earlier enterprise decision worth reconsidering?

When credible developments materially change the assumptions behind it, such as technical feasibility, supplier economics, customer attitudes, competition or regulatory requirements. A new model announcement or survey finding is not necessarily enough. The distinction is between responding to consequential evidence and repeatedly reopening commitments because the surrounding commentary has changed.

Originally published on Abhishek Sinha's personal site. This copy is shared on Epistamate Insights for readers interested in enterprise judgement, evidence and adaptation.

← Back to Insights