Lazy Fucking Interns

A month ago, a friend of mine was working through something with a member of his team. He’s a CEO, and this was a strategic decision – should they shut down this business line and use the resources to launch something new? The person on his team was senior and fully responsible for the decision – whether to shut down, how to iterate, the direction and scope of whatever came next. This employee sent my friend a slide deck explaining what the new business could look like. The email had no context, and the deck was very clearly AI-generated – it only reflected some of what the two of them had actually discussed in previous meetings. Left with a lot of questions, my friend sent back a long email full of feedback.

A couple of days later, an email back in response. Again, zero context – just “how about this?” with a link to a completely different slide deck, pitching a completely different business. No explanation of how he’d addressed the earlier feedback, or why this concept was better. No way to follow the logical thread that had gotten them here. It was like starting the conversation over from zero.

Now, all of us who work with AI know what happened here: this employee had a conversation with AI based on the previous meetings and discussions about the business problem. He asked the model, whichever one it was, to help him come up with an idea and generate a slide deck based on whatever parameters he fed it. Then, when he got feedback from his CEO, he fed that feedback into the same conversation and it spit back a new concept. Who knows how much he iterated or went back and forth with our robot friends on either deck, but what came out was a pretty useless waste of time for everyone involved, because it lacked judgment, context, and the connective tissue that makes these conversations move forward.

As Hilary Gridley pointed out recently, in the last year, the problem has moved from “how do I get people on my team using AI?” to “how do I get people on my team to stop using AI so badly??”

A lot has been written about the concept of AI slop, but I want to talk about this workflow and ones like it because I think under the label “AI slop,” there are a lot of different problems. I have taken to calling this one the “Lazy Fucking Interns” problem.

You know that behavior that used to drive you nuts when you were training interns? You’d sit an intern down, give them all the context you thought they needed and pretty clear instructions (in your mind, anyway) of what you needed, and then ask them to generate a first draft of a report or an email or a project. Once they were done, they’d send you an email that was one sentence – the equivalent of “here you go” with no further explanation – and attach a first draft that was missing detail and context, hadn’t been thought through all the way, and generally didn’t “show their work.” One of the hardest things about coaching interns, for me anyway, was explaining the gap between their definition of “good” and mine, their definition of “done” and mine. Interns often think the point is completing the project, not solving the larger problem. Progress for the sake of progress. Sometimes, it seems, they don’t even grasp what the larger problem is. Teaching someone judgment is exhausting. Sometimes the gap was so big I didn’t even know where to start with coaching them.

That same gap is exactly what’s playing out right now in organizations everywhere. There are essentially two entangled problems coming together to create a lot of stupid behavior and wasted time.

The first is that people are treating AI chatbots and agents as if they are highly skilled senior employees when, in most cases, they aren’t. They’re interns. Often sycophantic and kinda lazy ones.

AI’s job isn’t to think critically about what’s right at the highest level – it’s to deliver an answer. If that answer gets treated without a critical eye, without human judgment, the Lazy Intern vibes just get passed along. Just like interns, AI is great at useless productivity – shipping something regardless of whether it actually answers the question or meets any bar for quality. When someone copies and pastes the highly-confident-and-very-pretty-but-full-of-holes deck or email or doc that AI generated, you get a chain of Lazy Intern work.

And that’s exactly the second problem: somehow AI is making smart, skilled, senior people lazier. I can’t quite tell if it’s a pass-through effect, or if working with such a confident, seemingly smart intern like Claude or ChatGPT just makes us feel smart, so we’ve stopped asking questions. Either way, the downstream effect is that behavior that never would have been acceptable before – passing your manager a report an intern wrote with no significant editing or vetting – is somehow becoming commonplace.

The Lazy Intern pattern we’ve been talking about a lot in Glue Club is senior people sending over work with very little context that’s clearly AI-generated, and then, when you start asking questions about it, immediately naming or blaming the robots. “Oh, I didn’t write that section, that was Claude” or “Yeah, ChatGPT put that in.” As if there’s an autonomous third person in the room who can take the fall for being thoughtless, or… lazy.

We are managing Lazy Interns, and we are becoming Lazy Interns. Lazy Fucking Interns everywhere!

Ok so how do we combat this Lazy Intern problem? Some ideas: