Why the next decade of enterprise services will reward capability over capacity, and what that asks of the people who build this industry.
For most of my career, the IT services industry has created value in a way everyone understood. We brought skilled people, global delivery, process maturity, and the ability to execute at scale and at a sensible cost. When an enterprise needed capacity it didn't have, specialised skills it couldn't hire fast enough, or predictable execution across time zones, we were the answer. That model has served clients, shareholders, and millions of careers well.
I don't think it's dying. But I've become convinced it is quietly being repriced.
Enterprises are now adopting AI inside their own walls. They can see for themselves that work which once required large teams can be accelerated, automated, or augmented. So they are starting to ask their service partners a sharper question than they used to. Not "what does this cost per hour?" but something closer to:
What can you create that we can't easily create ourselves?
That is a harder question, and a fairer one. It pushes our industry away from models built on effort and toward models built on value. Not everything changes overnight, and I'll argue below that a lot of the breathless predictions are wrong. But the basis of differentiation is shifting, and leaders who pretend otherwise are setting their people up to be surprised.
Before the frameworks, it's worth being concrete about what an enterprise is actually buying. In practice, value in our world reduces to four things a client can measure.
A lower cost per unit of work, achieved through efficiency, automation, and reuse rather than cheaper labour or thinner margins.
More delivered for the same budget, because productivity expands output without expanding cost.
Faster releases, faster resolution, faster time to market.
Fewer defects, better reliability, stronger compliance, a better experience.
None of this is exotic. It's what enterprises already report getting from AI internally. In Deloitte's State of AI in the Enterprise 2026 survey of more than 3,200 leaders, improving productivity and efficiency topped the list of realised benefits, with roughly two-thirds of organisations reporting gains there.1 The implication for us is direct. If clients are seeing those gains on their own, they will expect us to bring at least as much into the work we deliver, and to share it, not pocket it.
I want to plant a flag here, because the temptation in a piece like this is to narrate an inevitable, frictionless future. That future isn't the one in front of us.
of agentic AI projects will be cancelled by the end of 2027, Gartner expects, citing escalating costs, unclear business value, and inadequate risk controls. Its analysts estimate that only about 130 of the thousands of vendors marketing "agentic" products are doing anything genuinely agentic. The rest are what Gartner bluntly calls agent washing.4
Gartner, June 2025
Deloitte's own data tells a similar story of ambition outrunning reality. While efficiency gains are widespread, only about a third of organisations are using AI to genuinely reimagine how they work, and revenue growth from AI remains far more hoped-for than realised.1
I read that not as a reason to slow down, but as a reason to be honest. The winners in this transition will not be whoever moves fastest or talks loudest. They will be whoever moves with discipline, and whoever is candid with clients about what AI can and can't yet do. For a services industry whose entire premise is trust, that candour is itself a competitive asset.
With that caution in place, here is how I see the four models that define our business changing.
How a services company creates value, where it delivers from, what it commits to, and how it bills for it. These four models define the shape of any IT services business. AI is redrawing each one, in different ways and at different speeds.
How a services company creates and captures value is the deepest layer, and AI does not touch all of it equally. Four flavours of the business model are moving at very different speeds.
Feels the most pressure. The honest client question is now unavoidable: if AI is lifting productivity, why am I paying for the same number of hours, and why should the productivity gain sit only with the vendor? This work doesn't vanish, but its defensibility erodes unless the talent is genuinely premium.
Becomes the most fertile ground. When you already own the running of a function, you have the standing to redesign it with automation, predictive operations, and agents, and to prove the improvement. AI shows up here not as a slide but as a lower cost per ticket and a faster mean time to resolution.
Grows in importance. A single retailer may not build its own store-operations engine; a partner serving fifty retailers can build the asset once and deploy it many times. The industry knowledge we've accumulated over decades is, in effect, an un-monetised library of reusable IP. AI is what finally lets us turn it into product.
Expands, but selectively. Committing to a measurable result only works where the baseline is clear, the outcome is attributable, the provider has real control, and there's prior proof it can be done. That describes a meaningful slice of work (service desk, claims, cloud optimisation, renewals), but not all of it. The discipline is knowing the difference.
I'll be direct about something I think gets over-predicted: AI is not going to erase onsite, offshore, nearshore, or hybrid delivery. Each location still has a job that doesn't disappear. What changes is how the work inside each location actually gets done.
Anchors governance, discovery, and transformation. Stays essential for work that needs presence, context, and stakeholder access.
Carries scale, specialised skills, and follow-the-sun coverage. Becomes an AI-augmented pod rather than a body-count pool.
Bridges time zones and regulatory geographies. Sits where speed of collaboration and proximity matter more than absolute cost.
Combines all three. Remains the default for large accounts, where transformation, run, and innovation happen in parallel.
The engagement model answers a simple question: what is the provider on the hook for? AI redraws each answer differently, and the result is a sharper distinction between commodity engagement and strategic engagement.
Survives, but generic staffing loses its shine. "Can you give me five Java developers?" becomes "Can you give me engineers who make AI tools productive, validate what those tools generate, and understand my domain?" Skill mix, not headcount, becomes the unit of value.
Splits in two. As internal teams get more productive, some enterprises will pull core build and transformation back in-house, and they should. What remains for partners is the specialised and non-core: network modernisation, compliance-heavy work, IP-dependent delivery where we bring playbooks and implementation scars the client doesn't have.
Becomes the most visible site of AI transformation. Gartner projects that by 2029, agentic AI will autonomously resolve 80% of common customer service issues, with an associated 30% cut in operational costs.3 The pitch moves from "we'll staff this process" to "we'll run this process with agents, automation, domain experts, and human supervisors, and show you the numbers."
Becomes more valuable precisely where the work is hardest: rethinking processes, operating models, and customer experience around AI. It's the one engagement model AI makes richer rather than leaner, because it's the one that depends most on human wisdom about where to point the machines.
Pricing is where the pressure becomes unavoidable, because AI breaks the old link between effort and output. Four pricing modes survive, but each one is being reshaped, and a fifth (verified outcomes) is rising fast.
Survives where flexibility and discovery matter: advisory work, early AI exploration, genuinely unclear scope. In mature delivery, billing purely for hours becomes hard to defend once everyone knows AI improved the productivity behind those hours. Expect T&M to go hybrid, with new units creeping in: platform usage, compute and token consumption, expert-review time, human-in-the-loop validation.
Stays, but its construction changes. From pricing human effort, to pricing a packaged offering that bundles accelerators, agents, frameworks, and experts. The client gets predictability; the provider earns margin from reuse rather than from bodies.
Grows wherever the service needs continuous improvement, which describes most AI-enabled operations. Models, agents, and governance all keep moving, and clients increasingly want ongoing capability rather than a one-time build.
Grows where proof exists. The clearest current signal is Zendesk, which now prices its AI agents on verified resolutions. The client pays only when an issue is genuinely resolved without a human, with a verification step deciding what counts.5 A long way from a per-seat licence, and a preview of where serious commercial models are heading: away from activity, toward audited value.
Everything above is a story about models. But our industry has never really been about models. It's about the millions of people whose work, and whose sense of themselves, is bound up in it. If I'm honest, that's the part of this transition I think about most, and it's the part the strategy decks tend to skip.
The work is changing under people who did nothing wrong. Someone who spent a decade becoming excellent at a task an agent can now do is owed more than a reorganisation chart. The leaders I respect are not asking "how few people can deliver this?" They're asking "how do we move our people up the value chain rather than out of it?"
The constraint on this transformation is human. And so is the advantage.
Encouragingly, the data suggests this is where the front-runners are already investing. In Deloitte's survey, the AI skills gap (not the technology) was named the single biggest barrier to integration, and the most common response among leading organisations was to upskill, reskill, and educate their broader workforce, with many redesigning career paths altogether.2 The firms that win the next decade will be the ones whose people learned to supervise, judge, and direct AI faster than their competitors' people did. That only happens in cultures that invest in them rather than discount them.
This is also, frankly, where humility belongs. None of us knows exactly which roles compound and which fade. What we can control is whether we bring people through the change with respect, clarity, and real investment, or whether we treat them as the line item that AI was supposed to reduce. I know which kind of leader I want to be, and which kind of company outlasts a hype cycle.
Strip away the frameworks and the through-line is short. AI is moving the centre of gravity in our industry from capacity-led value to capability-led value. From how many people we can supply to how much capability we can compound and prove.
The business model tilts toward IP, platforms, managed operations, and outcomes. Delivery stays global but becomes AI-enabled. Engagement moves from generic staffing toward specialised work and genuine partnership. And pricing moves from effort toward audited value.
AI will not make every services vendor more valuable. If anything, it will expose the ones still selling effort dressed up as transformation. The partners who earn the next decade will answer four plain questions without flinching.
Can you lower my cost per unit of work without lowering quality?
Can you deliver more within the same budget?
Can you get me to market faster?
Can you make the experience better?
Those who can will thrive in exactly the conditions that unsettle everyone else: when transformation is the work and uncertainty is the weather. That has always been the best argument for our industry's existence.
AI doesn't end it.
It just raises the standard, and asks us to bring our people with us.
The next decade will reward whoever builds capability, prices on proof, and treats their people as the source of the advantage rather than the cost of it.