The best AI lesson this summer came from watching our interns challenge AI

Every conversation about AI and early-career employees seems to start in the same place: will it weaken foundational skills by doing too much of the work? That's a reasonable concern. It's also not the question that ended up mattering most this summer.

Whether someone knows when to question the output turned out to be harder to develop than any of us expected, and more important than any other skill we could have focused on.

We made AI part of the work

Like many organizations, we're focused on using AI to innovate how we do everyday work. We didn't treat it as a separate training course or reserve it for experienced employees. We gave our interns access to the same tools our teams use and expected them to incorporate them into real projects from day one.

One intern built a company-wide AI prompt library that's still used across the organization today. Others relied on AI while reviewing contracts, planning communications projects, and working through engineering challenges.

The work they produced was impressive. What stayed with me wasn't the speed or the quality of the first draft. It was watching them decide what to keep, what to change, and what to ignore.

The hard part started after AI answered

Although they worked in different functions, each intern eventually reached a similar point.

Kahari, our legal intern, found that AI could quickly surface relevant precedent, identify issues in contracts, and draft language. Choosing which legal argument was actually stronger still depended on understanding the reasoning behind it.

Constanza, who joined our communications team, used AI to draft employee communications, evaluate internal communications platforms, build project plans, and create repeatable workflows. More than once she rewrote AI-generated drafts because they were technically accurate but missed the tone or context the situation called for.

Jennifer, working on an engineering project, used AI almost every day to understand unfamiliar infrastructure, validate technical approaches, and automate repetitive tasks. She described it as something that saved time while helping her think more creatively about the problems she was solving. Even then, she still had to decide whether the suggested solution actually fit the system she was building.

Those moments had little to do with the technology itself. They came from conversations with teammates, feedback from managers, and gradually building enough context to recognize when something that sounded convincing wasn't the best path forward.

The learning didn't disappear. It shifted.

One concern I hear often is that AI will remove the struggle that helps early-career employees develop expertise. The technology removed work that rarely teaches someone much in the first place: searching for background information, getting oriented in unfamiliar systems, producing a first draft. The work that mattered stayed exactly where it has always been: making decisions, defending them, responding to feedback, and improving the next version.

AI is already part of how younger employees work. Deloitte's 2025 Gen Z and Millennial Survey found that more than half of Gen Z respondents already use generative AI at work, and nearly three-quarters expect it to reshape how they work in the near future. The question for employers is no longer whether early-career employees will use AI. It's whether we'll give them the coaching, context, and accountability they need to use it well.

Teach people how to think, not just how to prompt

Learning to write a better prompt is useful. Learning to recognize when an answer deserves another question is far more valuable, and that comes from experience, feedback, and taking responsibility for work that matters.

The prompt library our intern built wasn't the most significant outcome of the summer. What I'll carry into next year is watching our interns grow more confident as they took ownership of work with real consequences. AI helped them get started faster. The decisions that mattered were still theirs to make, and accountability is the part no tool changes.