AI Strategy in 2026: How AI Is Changing the Economics of Business

By Caroline Kennedy | Published October 08, 2026

 

For the past few years, most conversations I've had with leaders about AI have started in roughly the same place: productivity. Where can we save time? What can we automate?

Those are still useful questions. But the conversation is moving somewhere more consequential, and it starts with a number that caught my attention.

 

Why AI Strategy Is Becoming Business Strategy

McKinsey's State of AI 2026 report found that 32% of organisations have already decided against purchasing at least one software product or feature because they could build the functionality internally using AI coding tools. In technology companies, that figure rises to 41%.

That's no longer about writing an email faster or analysing a spreadsheet in half the time. It's starting to affect what businesses buy, what they build themselves, where technology budgets go, and which capabilities need to sit inside the organisation.

Once those economics shift, the effects don't stay contained. A capability that used to require another software licence can now be built in-house, which changes the buying decision. A team that can suddenly handle more work doesn't need the same structure it had before. And as spending on AI rises alongside pressure to control token and operating costs, leaders need a clearer answer to where that investment actually creates value.

None of that is a technology question at its core. It sits much closer to the operating model of the business, and that's where I think the next stage of the AI conversation needs to go.

There has always been a trade-off between building something yourself and buying it from someone else. A business identifies a need, leadership weighs the options, and building internally usually means more control at the cost of people, money and time.

For most of the past decade, the simpler answer won: buy the software. SaaS made that decision even easier, letting organisations subscribe to specialist platforms instead of maintaining the technology themselves.

AI is complicating that logic. EY is seeing a similar shift from the executive side. Its 2026 AI Pulse Survey found that 87% of senior leaders investing in AI say their organisations have either deployed or are piloting programmes to build AI software internally, and 94% say AI lets them build software faster than traditional development allows.

If companies can create useful internal applications faster and more cheaply, some of the old assumptions behind enterprise software purchasing start to wobble. EY found that 95% of senior leaders expect their relationship with traditional software vendors to change within five years, and 82% believe per-seat SaaS pricing will become less relevant in their industry.

I don't read this as the end of enterprise software, and it certainly doesn't mean every company should start building everything itself. What it points to is something more interesting: the line between what belongs inside an organisation and what should be bought from outside it is becoming less fixed.

 

Why AI Budgets Are Harder to Manage Than Software Budgets

If AI is changing what businesses buy and what they can build, the next question is where the money should go, and that's turning out to be harder than it sounds.

McKinsey found that one in five organisations has already constrained its use of AI because of operating costs, including tokens. Yet businesses aren't retreating. Sixty per cent expect to increase their AI investment over the next year, and 28% are already putting more than 10% of their enterprise technology budget toward AI.

That tension is worth sitting with. The question isn't really whether AI is worth investing in anymore. It's whether leaders understand the economics of what they're investing in.

Traditional enterprise software gave leaders a familiar cost structure: licences, seats, subscriptions. AI works differently. Costs can move with usage, model choice, processing demands and how deeply AI gets embedded across the organisation, and as adoption spreads, that becomes harder to track.

KPMG's latest Global AI Pulse found that 42% of leaders have only partial visibility into AI spending, 23% are struggling with usage-based costs, and a third cite limited understanding of AI cost structures as a challenge when deploying AI agents.

The more telling number is what happens once organisations do understand those costs. Those with strong visibility into AI operating costs were five times more likely to report established ROI: 15%, versus just 3% among organisations without that visibility.

That tells me something about how leaders need to think about AI now.

 

Where Does the Real Return on AI Investment Come From?

If AI investment is a capital allocation decision, there's an uncomfortable question underneath it: what are you actually investing in? More tools, more automation, more licences, or a genuinely different way of operating?

McKinsey's research offers a useful clue. Only 6% of respondents qualify as AI high performers, meaning organisations reporting significant value from AI and attributing at least 5% of EBIT to it. What separates that small group isn't that they use more AI. It's what they're willing to change around it.

Nearly three-quarters of AI high performers have fundamentally redesigned workflows because of AI, compared with roughly a quarter of everyone else, and that share has climbed sharply from 55% last year among high performers.

That's one of the more important findings in the whole report, because most organisations find ways to slot AI into work that already exists. The organisations getting stronger results seem more willing to go back to the beginning and ask whether the workflow still makes sense at all: why a step exists, why a particular person needs to touch it, and what changes once AI can do part of that work instantly.

Maybe that's the real dividing line: some organisations are using AI to make the existing business run faster, and others are using it to ask whether the business should run the same way at all. If that’s the case then it's a question of workflow design, roles, capacity and the operating model itself, which raises an obvious next question: what does this mean for the people doing the work?

 

Will AI Actually Cut Jobs?

McKinsey's latest findings suggest expectations are shifting fast. Thirty-nine per cent of respondents expect AI to reduce their organisation's overall headcount over the next year, while 43% expect little or no change.

There's a good reason to treat those numbers carefully. In McKinsey's 2025 survey, 32% expected AI-related workforce reductions the following year. When asked what actually happened, only 14% reported that AI had contributed to an overall decline in their workforce. The anticipated impact ran at more than twice the realised one.

That doesn't mean AI won't affect employment. It suggests the effect is more complicated than simple replacement.

PwC's 2026 Global AI Jobs Barometer adds an interesting counterpoint. After analysing more than one billion job advertisements across six continents, PwC found that companies most able to use AI had experienced 52% headcount growth relative to 2018 levels, compared with 36% among the least AI-exposed companies.

At the same time, the skills in demand are shifting. PwC found AI-exposed work leans harder on judgement, creativity and leadership. In its US analysis, highly AI-exposed entry-level roles were seven times more likely to require skills traditionally associated with senior employees, including leadership and strategic decision-making.

So the workforce question for leaders probably shouldn't stop at "how many people will AI replace." It's closer to: what work no longer needs doing, what new work becomes possible, and where does human judgement become more valuable as AI absorbs more of the routine work?

 

Why Every AI Decision Eventually Becomes an Operating-Model Decision

A leader might start with what looks like a contained AI decision. Maybe the organisation discovers it can build a capability internally instead of buying another software product. That changes the technology budget, which then changes the workflow once the new capability automates part of an existing process.

That's where leadership faces another decision: reduce that capacity, redirect it, hand people different responsibilities, or point it toward customers, growth or work the organisation never previously had time for.

Follow that chain far enough and what started as a technology choice becomes a question about how the business operates. That's why I think "AI strategy" is a slightly misleading phrase. It can make AI sound like a separate programme sitting next to the rest of the organisation.

Organisations that are getting stronger results are doing close to the opposite. Their AI initiatives connect workflow redesign, workforce planning, cost management, impact measurement and senior leadership ownership, and those pieces reinforce each other.

You can't fundamentally change a workflow without thinking about the people inside it. You can't change roles without thinking about capability. You can't keep increasing AI investment without understanding the return, and you can't make those calls independently across technology, finance, HR and operations and expect a coherent transformation to fall out of it.

The limiting factor is increasingly the organisation's ability to absorb change, and I think that's worth sitting with. The technology will keep getting better. The harder question is whether the organisation can change alongside it, whether leaders can make decisions across functional boundaries, remove processes that no longer make sense, and redesign roles before people simply inherit more work.

 

Lesson: What Kind of Business Are You Actually Building?

There's a difference between putting new technology into an old business and redesigning a business around what that technology now makes possible.

Think about renovating a house. You can replace the appliances, install smarter lighting and buy better furniture while leaving the floor plan exactly as it was. Everything gets a little faster and easier, but it's still fundamentally the same house.

That's how a lot of organisations are approaching AI right now: dropping it into existing jobs, workflows and structures. The report gets written faster, the meeting notes appear automatically, the customer response takes seconds instead of minutes. Those improvements matter, but at some point leaders have to ask whether they're just upgrading the appliances while leaving the floor plan untouched.

Value doesn't appear automatically because the technology improves. Someone has to redesign the system around the new capability, and that's ultimately why AI strategy is becoming inseparable from business strategy. The decisions start with technology, but they quickly reach the foundations of the organisation: what you own, what you outsource, where you invest, how work moves and where you want the business to create value.

So maybe the job of leadership isn't bolting AI onto the organisation that already exists. It's looking at that organisation with fresh eyes and asking what to keep, what to change, and what to stop building the same way.

That's a much bigger conversation than AI adoption, and I think it's the one topic senior leaders now need to be having.

 

About Caroline Kennedy

Caroline Kennedy is an award-winning former CEO, CEO and executive coach, and keynote speaker who has led multinational businesses generating up to $250 million in annual revenue. She works with CEOs, executive teams and organisations on leadership, business performance, transformation and the practical application of AI, helping leaders understand what emerging technology makes possible and how to turn that possibility into meaningful business value. If you’re ready to become a more effective leader and create better conditions for your people to perform, find out more about Caroline’s Executive Coaching.

 

Sources:

McKinsey & Company. (2026, August 25). The State of AI in 2026: On the Road to ROI.
https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai

EY. (2026, July 28). C-suites pivot from AI adoption to unlocking value as escalating token costs trigger fiscal scrutiny.
https://www.ey.com/en_us/newsroom/2026/07/ey-survey-c-suites-pivot-from-ai-adoption-to-unlocking-value-as-escalating-token-costs-trigger-fiscal-scrutiny

KPMG. (2026, June 24). Growing adoption signals progress as cost visibility and accountability drive AI value.
https://kpmg.com/xx/en/media/press-releases/2026/06/growing-adoption-signals-progress-as-cost-visibility-and-accountability-drive-ai-value.html

PwC. (2026, June 15). 2026 Global AI Jobs Barometer.
https://www.pwc.com/gx/en/issues/artificial-intelligence/publications/artificial-intelligence-study.html