November 20, 2026

How to prepare your existing engineering talent for AI-driven change

Lucy-Suslova-Head-of-Engineering-Excellence-at-Intellias

AI is changing how software gets built and what companies need from their engineering teams. As AI takes on more of the implementation work, engineers need to do more than write code. They need to understand business problems, guide AI agents, and verify their output. For companies, the challenge is how to help existing teams develop these skills while keeping up with delivery.

But the re-skilling process requires more than launching new tools. It means rethinking the engineering role, the way teams learn, and how companies prepare people for a different delivery model.

AI has already been redefining the engineering role

Less of the day goes into writing code. Engineers explain what they want, set up context for the agent, break work into pieces it can handle, and review what comes back. That’s a different kind of tiring: less typing, more judgment. They read far more code than they write, and a lot of it looks correct whether or not it is. From what I’ve seen, the artifacts that matter most are now the spec and the tests the engineer owns, more than the pull request.

I’m biased because I came into management through quality engineering. I believe verification is now one of the fundamental engineering skills. Every engineer is now a reviewer of code they didn’t write, in quantities that no one ever prepared them for. The scary thing is that AI mistakes can be absolutely plausible. That’s why context engineering and evals are so important and why you need to understand enough about what’s going on under the hood to be able to see why an output went wrong. Systems thinking is the thing that brings it all together because generated code that completely ignores your architecture is just expensive noise.

Besides technical skill, engineers need to know when to stop. When to take the wheel back from the agent and when “good enough” really isn’t. The best engineers I work with also change their minds fast when the evidence shifts, whether that bad idea was theirs or the model’s. Then there’s product sense, that gut feeling for what the user actually needs, which you only get from talking to the customer. And the ability to explain a trade-off to a business person without all the jargon is quietly becoming a career multiplier.

Why companies need to re-skill existing engineering talent

There are several reasons why re-skilling matters so much nowadays. The first is context. An AI agent in your codebase is only as smart as the context you feed it. Who has that context? The people who’ve been working inside those systems for years and still remember why one particular service falls over every December. New hires bring fresh perspectives and valuable expertise, while existing team members bring a deep understanding of the business and its systems. Re-skilling brings these strengths together, so AI becomes part of how the wider engineering team works, rather than a separate group’s responsibility. 

The second reason is business risk. Buying AI tools and running a few pilots doesn’t in any way mean that the way you deliver software is going to change. Pilot fatigue comes first. You’ve bought the licenses, you’ve had a nice demo from someone, but two years down the line the delivery is still the same, and the CFO is starting to ask where the money’s gone. Shadow AI is worse because you don’t see it. People without approved tools will just take your code and paste it into whatever is open on their laptop. 

Re-skilling helps companies turn AI investment into a real change in how they work. Once code gets cheap, the expensive bit is deciding what to build and proving it works. A re-skilled team can try a few approaches in the time one used to take and kill the weak ones before they sneak onto a roadmap. What’s more, your competitor can buy the same licenses as you have tomorrow morning. What they can’t buy is a team that knows where to aim all that new capacity and AI harness.

Building an AI-ready engineering team

How do you adapt a team to the new AI era? It starts with two mindset shifts. The first is going from deterministic to probabilistic thinking. Engineering profession grew up on “same input, same output,” and AI simply doesn’t follow that rule. You stop expecting correctness and start designing for it, with evals and guardrails.

The second shift is personal, and people avoid talking about it. A lot of senior engineers built their identity on beautiful handwritten code. Admitting that part of the craft is becoming a commodity hurts a bit, and the ones who adapt fastest shift their pride to the outcome. Learning also adds to the delivery load, and experienced engineers may fear looking like beginners in front of their teams. Companies need to address these barriers by providing approved tools, clear security rules, and a sandbox for experimenting with realistic code.

One more important thing: learning needs to reflect how quickly AI is changing. That’s why I’d run re-skilling as a product rather than a course: it’s versioned and updated every few months, since the content goes stale fast. Starting with a shared foundation, then going deeper into what people actually do, with the practitioners themselves showing their own failures instead of vendor slides. Training also needs to lead back into delivery. The bit most programs skip is a real project where people can put what they’ve learned, supported by a manager who expects to see new practices in use. Otherwise, teams will keep working the old way.

Putting re-skilling into practice

In my experience, effective re-skilling requires combining engineering and AI skills with product thinking and business understanding. This shift is needed because I see that clients are starting to ask how an AI-enabled team can take a problem from idea to production. Previously, learning paths focused on engineering depth and breadth, but AI blurred the boundaries between them. It became clear that engineers needed to connect technical execution with a deeper understanding of what needs to be built and why.

This isn’t about turning engineers into product managers. It’s about engineers who know why something is worth building and can use AI to build it well. That means understanding the client’s problem, knowing how to work with AI agents and evaluate their output, and using AI throughout the delivery lifecycle, not just to write code. The goal is to help engineers take an ambiguous client problem all the way to production.

From what I have observed, this approach helps build stronger AI-enabled teams and develop practices that can be reused across projects. But how do we know if it works? I’d borrow metrics from the growth playbook, because AI adoption has a lot in common with product adoption:

  • Retention: Are the practices still being used three months later? Has another team picked them up?
  • Delivery outcomes: How quickly does an idea become a validated prototype? Does quality hold up as teams move faster?
  • Ownership: Are engineers taking greater ownership of client problems?
  • Manager feedback: Does the manager see the same progress as the engineer? This matters because confidence can grow far faster than capability.

More broadly, the goal is to build a habit of adapting, not just a set of AI skills. The next big shift may not be about AI at all, and teams will be better prepared for it if engineers already know how to rethink the way they work.