AI Doesn't Shrink IT Vendors. It Sorts Them. 

Why traditional IT services companies need to rethink where they create value?

Almost every enterprise I speak with is running some version of the same experiment: give AI to the engineering team and measure what happens to productivity, cost and delivery speed. 

For enterprises, that is a sensible question. For IT services companies, however, I think there is a more important one: if AI allows the same team to produce significantly more, what will clients still pay us a premium for? 

For much of the past two decades, the economics of IT services were closely tied to engineering capacity. Clients had technology needs; vendors supplied people with the skills and time to execute them. More work generally required more people, which made hours and headcount a reasonable proxy for both cost and value. 

AI is beginning to weaken that relationship. 

This does not mean software engineering is becoming easy. Enterprise systems still involve difficult architecture, legacy environments, integrations, security, compliance and production risk. But many activities within software delivery are becoming faster. An experienced engineer working effectively with AI can already cover considerably more ground than the same engineer could a few years ago. 

As AI-driven productivity becomes more common, clients will naturally expect to participate in those gains. In the short term, a services company may capture higher margins by using AI to deliver the same work with less effort. But that advantage is unlikely to remain proprietary. Competitors will adopt similar tools, clients will adjust their expectations, and eventually the productivity gain will show up in some combination of lower cost, shorter timelines or greater scope. 

This does not make engineering execution unimportant. Reliable enterprise delivery remains difficult, and strong engineering capability will continue to matter. What changes is its ability to differentiate a services company on its own. When AI-enabled productivity becomes part of the market baseline, simply being able to execute requirements efficiently becomes less distinctive. The question then becomes what additional value a technology partner brings around that execution - in judgment, architecture, domain understanding, risk management and ultimately business outcomes. 

What becomes more valuable when execution gets faster? 

My answer is judgment. 

Someone still needs to determine whether we are solving the right problem, whether the requirements reflect how the business actually operates, whether an architecture will survive real-world scale, which technical debt is acceptable, and when an AI-generated solution that looks convincing is actually wrong. 

These decisions have always existed. Historically, however, their value was less visible because implementation consumed so much of the effort, time and budget. AI changes that balance. As execution becomes faster, the quality of the decisions directing that execution becomes more consequential. 

If AI can draft user stories in minutes, the difficult part is no longer producing them, but knowing whether they represent the real business problem. If AI can propose architectures, the value lies in recognizing which assumptions may fail in production. And as AI generates more code, experienced engineers spend less of their value producing syntax and more of it deciding what should be accepted, challenged or redesigned. 

This creates an important paradox. We often describe AI as a labor equation: if AI does more work, fewer people are required. But there is another equation that matters just as much: as AI increases execution leverage, it also increases the leverage of human judgment. 

One decision can now influence far more downstream output. When that decision is sound, the productivity gain can be enormous. When the underlying assumption is wrong, AI can scale the mistake just as efficiently. In that sense, AI increases the blast radius of human judgment. 

This is why I don't believe the future delivery team is simply today's team with fewer people. A smaller AI-enabled team may produce substantially more, but its decisions also carry greater consequences. That increases the importance of product judgment, architecture, quality, risk ownership and accountability for outcomes. 

AI can help a good team move much faster. It can also allow a poor decision to scale much faster. The opportunity for IT services companies is therefore not simply to reduce human effort, but to put the strongest human judgment where AI creates the greatest leverage. 

Clients ultimately buy outcomes, not hours 

When I look beyond the language of RFPs (Request for Proposals), I think enterprise clients ultimately expect a technology partner to help them achieve three things: make fewer expensive mistakes, make important decisions faster, and create better business outcomes. 

None of these is purely an execution-speed problem. 

A team can produce code three times faster and still build the wrong product. It can deliver every sprint on time and still create an architecture that becomes expensive to scale. It can meet every requirement in the original specification and still fail to improve the business metric that justified the investment. 

This is where I believe the traditional boundary between “consulting” and “delivery” starts to become less useful. A technology partner cannot take responsibility for outcomes if it only enters the conversation after the important decisions have already been made. 

That pushes services companies upstream - into discovery, product thinking, architecture, domain understanding and risk, and downstream, toward measuring whether what was delivered actually created value. 

This does not mean every IT company needs to become a consultancy 

There is an important counterargument. 

If AI makes engineering significantly more productive, some companies may succeed precisely by becoming extremely efficient AI-native execution businesses. Others may differentiate through deep platform expertise, cybersecurity, managed services, industry specialization or integration capability. 

I don't think there will be one winning model. 

AI may also create more technology demand rather than simply reduce engineering demand. As the cost of building software falls, projects that previously did not justify the investment become economically viable. We may therefore see more software being built, while requiring fewer people per unit of software. 

That distinction matters. The market does not have to shrink for the traditional outsourcing model to come under pressure. 

The change may instead be a repricing of value: less premium for undifferentiated capacity, and more value accruing to capabilities that remain difficult to replicate - domain knowledge, architecture, judgment, trust, risk ownership and the ability to connect technology decisions to business outcomes. 

What this means for us at Enouvo 

This is not an abstract question for me. Enouvo has been building software for 15 years, and we have grown through many of the same models as the rest of the outsourcing industry. We are now having to challenge some of the assumptions that helped build that business. 

One conclusion we have reached is that becoming AI-native cannot simply mean giving developers/ engineers better AI tools. The operating model has to change with the productivity model. 

We are increasingly organizing engagements around five accountabilities: AI-aware Product Management, architecture connected to business risk, senior engineering judgment, Quality & Risk ownership, and Delivery Leadership accountable for outcomes rather than only project metrics. 

These are not five new departments, nor necessarily five different people. They represent a change in what we expect our people to own. 

For example, we increasingly think about senior engineering less as a proxy for coding speed and more as a proxy for judgment. AI can provide much of the acceleration. What we need from experienced engineers is the ability to recognize when the accelerated answer is wrong. 

Similarly, Product Management becomes less valuable when it is limited to documenting requirements and tracking scope - activities AI can increasingly assist with - and more valuable when it challenges whether the requirements solve the right business problem. 

And quality can no longer mean only “does the software work?”. It increasingly needs to include whether AI-assisted development has introduced security, maintainability, compliance or architectural risks that functional testing may not reveal. 

We are still learning how to make this model work. I don't think anyone has completely solved the AI-native services company yet. 

But we have made a deliberate choice about where we want Enouvo to compete. 

We don't believe our long-term differentiation should come from providing more engineering hours at a competitive rate. We want it to come from combining strong engineering capability with enough product, architecture, domain and risk judgment to take greater responsibility for the client's outcome. 

The market is being repriced

Worldwide technology spending continues to grow, even as AI increases productivity. That suggests a more interesting future than the simple narrative that AI will eliminate large parts of the IT services industry. 

I expect two forces to operate simultaneously. 

AI will reduce the effort required to produce many forms of software. At the same time, lower production costs will make more technology investments economically viable. 

The result could be substantially more software, built with very different economics. 

That is why I don't think AI simply shrinks IT vendors. 

It sorts them!

Companies built primarily around selling capacity will need to decide whether greater efficiency alone is enough to sustain their differentiation. Others will move toward specialization, proprietary capability or deeper ownership of technology and business decisions. 

For Enouvo, our choice is to move from being primarily a capacity provider toward becoming a value partner. 

That transition is harder than adopting AI tools. It requires us to understand the client's business, challenge assumptions, make better decisions, accept more accountability and prove value beyond hours delivered. 

But perhaps that is the larger opportunity AI is creating for the IT services industry. 

For years, we have talked about wanting to become strategic partners to our clients. 

AI may finally force us to prove it!