Is AI Revealing the Gaps Teams Used to Hide? Chief Data Scientist Christopher Wells on Finding the Right Leaders to Fix Them

At a recent company meeting, the CEO did a bit of show-and-tell for the engineering team. Over a weekend, he had used AI to rebuild part of their product. It ran, it looked coherent, and it made the point he wanted to make, mainly that they should be moving faster.

But when he shared the code with the team, it didn’t hold up. The system had taken shortcuts, filled in gaps, and made decisions no one had explicitly defined. What came back looked right on the surface, but it wasn’t something the team could build on or meaningfully extend.

He set out to demonstrate what the tools could do. Instead, he showed where judgment, context, and experience still matter. Christopher Wells often sees versions of this. The issue isn’t technology. It’s how the work is defined. For PE-backed companies under pressure to move faster, improve margins, and tell a credible AI story, that gap is becoming clearer.

You’re managing work, not installing software

Wells started in theoretical physics before moving into finance as a quant, and eventually into enterprise AI across product and data science. Today, he is working on moving from pilot purgatory to something that holds up in the business.

What he sees is consistent. Teams focus on what the tools can do. Far fewer spend time clarifying the problem itself.

Working with AI, particularly agents, starts to feel less like using software and more like managing a junior team. They need direction. They lose context. They sometimes do something confidently wrong.

“The key skills for working with AI agents are basically classic management skills,” Wells said.

That shift exposes something that was always there. Many organizations rely on experienced people to fill in gaps. They apply judgment, challenge assumptions, and quietly correct things that don’t quite make sense. Over time, that becomes invisible.

AI doesn’t compensate in the same way. It reflects whatever clarity, or lack of it, sits behind the task.

Where things get messy

The day-to-day experience of working this way is also different from what most teams expect.

Wells described having multiple agents running at once when one of them interrupted him to announce it had finished a task. “It actually scared my kids,” he said. “It just started talking from my laptop.”

It’s a small moment, but it captures the reality. You’re not just issuing prompts. You’re managing attention, deciding what matters, and filtering a constant stream of output. “You can burn yourself out managing the herd.”

There’s also a quieter dynamic. These tools are rewarding to use. They produce output quickly, often convincingly, and sometimes better than expected.

“I get tons of endorphins from seeing something complete,” Wells said.

That can pull teams into an endless loop of generating and refining output without getting closer to a better answer. The work feels productive. It isn’t always.

At the same time, the gap between strong and weak operators becomes more visible. “The good people get really good, really fast,” Wells said. “And the bad people generate garbage faster than ever before.” What looks like a technology gap is often an execution gap.

What leaders are getting wrong

Most large organizations have responded to AI with policies. What not to do. What data not to use. Where to be careful.

Wells sees a different problem.

Companies aren’t investing enough in helping people get good at this. The gap isn’t access to tools. It’s capability.

Give people permission to use them. Show them how to use them well. Put real examples in front of them. Without that, companies don’t get leverage. They get noise, inconsistency, and little that translates into real performance.

That same pattern shows up in how companies approach implementation. A lot of early efforts start with a big question: what can we fully automate?

That’s usually the wrong place to begin.

The systems work better when the scope is narrow and the objective is clear. “You need a release valve,” Wells said. “The agent’s going to get stuck on something you didn’t plan for.”

In practice, the gains come from supporting work, not replacing it.

A clearer view of teams

“The number of jobs isn’t going down,” Wells said. “But the number of job titles is getting compressed.”

As AI takes on more structured work, roles blur and teams get leaner. The advantage shifts to people who can move across problems, ask better questions, and direct work effectively.

AI doesn’t change organizations as much as it reveals them.

Well-run teams get leverage. Poorly defined ones get exposed.

The difference isn’t access to the tools. It’s how clearly the work, and the people doing it, have been defined.