A version of this article is also available in Portuguese.

 

I have spent the last three decades building a technology services company and have survived every wave that was supposed to make us irrelevant.

“ERP will standardize everything.”
“Cloud will automate everything.”
“Digital transformation will be a destination: cross the finish line, declare victory, go home.”

Each time, the prediction was the same: the work would go away. And each time, the opposite happened.

So when I hear that AI will ‘finally’ kill IT services — that machines will write the code, design the systems, run the platforms, and relegate services firms to irrelevance — I don’t dismiss it intellectually.

The claim deserves to be taken seriously. Experience and past success don’t grant immunity from error. I have been wrong before about plenty of things. But I have also watched this exact narrative collapse under its own assumptions, repeatedly, for decades.

Let me tell you what I think is actually happening. And why it matters.

 

Every wave kept its promise, and that was the problem

Here’s the part where the ‘AI kills services’ story gets fundamentally wrong: it assumes the last major technology waves failed to deliver on their promises. They didn’t. They delivered. And by delivering, they created entirely new categories of hard.

Enterprise software delivered. It gave organizations integrated platforms, standardized processes, global visibility. Beautiful. But it also made something brutally visible: how genuinely different companies are from one another. Across industries, geographies, regulations, cultures. ERP didn’t fail to standardize; it succeeded in revealing that standardization was never the real problem. The real problem was the thousand undocumented decisions buried in spreadsheets, workarounds, and regional logic that existed for very good reasons. Software couldn’t adjudicate those differences. People had to.

Cloud delivered. No hardware, no long planning cycles, elastic resources on demand. Exactly as promised. And then infrastructure became something that never stopped changing. Architecture decisions multiplied. Cost management became a daily discipline. Security shifted from building a wall to playing an infinite game. A single misconfiguration could expose millions of records overnight. An unoptimized database could generate six-figure monthly bills before anyone noticed.

Digital transformation delivered. And then immediately proved there was no finish line. Every improvement reset expectations. Customers demanded more personalization. Regulators required more transparency. Competitors adopted new capabilities faster than any roadmap could anticipate.

The pattern is always the same: powerful tools don’t simplify the world. They give organizations the confidence, and the pressure, to attempt far more ambitious things. Complexity compounds. The work evolves. And the need for people who can navigate that complexity doesn’t shrink. It deepens.

 

Now it’s AI’s turn. And the stakes are higher.

I won’t pretend I’m not uneasy. Anyone running a services company who tells you they’re perfectly comfortable through this transition is either not being candid or not paying attention. AI changes things in ways that are genuinely discontinuous. The speed. The breadth. The sheer volume of what becomes possible overnight.

But here’s what keeps getting lost in the noise — and why the direction of this industry matters.

When something becomes easy, it becomes easy for everyone.

If AI can generate code, configure infrastructure, and spin up applications in minutes, those capabilities stop being differentiators. They become table stakes. Instantly. The competitive frontier doesn’t stay where it was. It moves. It always moves.

And where does it move? Toward the hard stuff. The stuff that can’t be generated. Integrating AI into legacy environments full of exceptions and technical debt. Embedding it into real operating processes. Not demos, not pilots, but the messy reality of how organizations actually work. Governing it under regulatory, security, risk, and ethical constraints that are only getting more complex. Running it reliably at scale, over time, with real accountability when things go wrong.

These are not problems AI eliminates. They are problems AI creates more of.

 

Better tools don’t simplify enterprises. They move the frontier.

This is the deepest misconception in the whole debate. There’s this assumption that if you give people better hammers, they need fewer carpenters. History says the opposite. Give people better hammers and they start building cathedrals.

When AI generates code, someone must still validate that it’s secure, compliant, and maintainable — and stand behind that decision. When AI proposes an architecture, someone must evaluate it against organizational constraints, accumulated technical debt, and a future no model can predict. Accountability does not get automated.

AI generates options. Judgment determines outcomes.

And judgment, real judgment, the kind that accounts for context, politics, risk appetite, regulatory pressure, legacy constraints, and human behavior, that doesn’t get automated. It gets more valuable. Because there are now far more options to judge.

Automation doesn’t eliminate the need for expertise. It shifts expertise upward. From execution to design. From implementation to judgment. From components to systems. AI commoditizes the easy parts and forces everyone to confront the difficult, interconnected, and often deeply uncomfortable work where real competitive advantage lives.

 

The real opportunity most people are missing

Here’s the part that genuinely excites me, underneath all the anxiety.

AI is a fundamentally new layer of abstraction in computing. Just as high-level languages expanded what engineers could build, and cloud expanded what organizations could deploy, AI expands the entire class of problems we can now address with software. Problems that were previously too expensive, too complex, or too slow to solve. Suddenly within reach.

That changes the ROI equation in ways we’re only beginning to understand. Experimentation gets cheaper. Iteration gets faster. The ceiling on impact goes up.

But, and this is critical, more possibilities doesn’t mean fewer decisions. It means more opportunities, more parallel initiatives, and exponentially more ways to get it wrong.

We’re already seeing it. Organizations aren’t asking for less help. They’re asking for a different kind of help. Not “build this for me” but “help me figure out what to build, in what order, and why.” 

That’s not the end of services. It’s the evolution of services toward where the real value has always lived.

 

The gap that AI widens

A fundamental mistake in the “AI kills services” narrative is treating capability as a substitute for judgment.

AI gives organizations—and their competitors—the same expanded set of options: faster execution and lower barriers to experimentation. What it does not provide is clarity. Which options matter. How to sequence them. How to align what’s technically possible with what the business actually needs.

That gap—the ambiguous space between what is technically possible and what is organizationally achievable—is exactly where we live. It’s where experienced partners earn their relevance by learning faster and accumulating asymmetric understanding across organizations, contexts, and failures.

And AI doesn’t close that gap. It widens it. Every new capability, every new possibility, every new tool stretches the distance between “we could do this” and “we should do this, in this way, right now.”

 

Selection, not collapse

This transition carries a hard truth: not every services firm will survive. Some business models are built on labor arbitrage and predictable, repeatable work that AI will genuinely displace. Companies that can’t move up—from execution to design, from delivery to partnership, from building what’s specified to helping clients figure out what’s worth building—are in real trouble.

But that’s not the death of an industry. That’s selection. It’s the same dynamic we saw with agile and digital, with cloud, and with every wave before it. The firms that couldn’t adapt didn’t prove the work was gone. They proved they couldn’t evolve with it.

At the same time, this is a huge opportunity. As AI expands what organizations can attempt, it expands the scope of the problems worth solving—and, with it, the potential size of the services market itself. The industry doesn’t shrink; it reorganizes around higher-value work.

 

History rhymes more than it breaks

Technology transitions are never clean. They unfold unevenly, over long periods, producing hybrids rather than replacements. Predictions of sudden obsolescence consistently underestimate how deeply technology becomes entangled with organizational, economic, and human systems.

AI will reshape how services are delivered, priced, and valued. It will raise expectations. It will compress timelines. It will demand that firms like ours combine genuine technical depth with something harder to replicate: deep contextual understanding of how organizations actually change.

What it won’t do is eliminate the need for experienced partners who can navigate complexity as it grows—and shifts.

The people predicting the imminent death of IT services are betting against the rhymes of history. I’ve been in this industry long enough to know. That’s rarely a winning trade.

The menu has changed. Again. The question is the one it’s always been: what do we build from here?

Cesar Gon is the founder and CEO of CI&T.