Letters page: My colleagues are producing AI workslop. How do I stop it?

"Dear Rich,

Three colleagues, including one of my most trusted direct reports, recently handed me a 40-slide deck. It's nicely formatted but I keep trying to read it and my brain just cannot do it. It's overly verbose, and I'm certain AI wrote most, if not all, of it.

Our company narrative is to embrace AI tools, especially Copilot. But this is becoming common. I now spend far longer reviewing documents than I ever have before, and I don't feel I'm gaining anything from it. I'm starting to feel like AI has become the first port of call, and that I might as well skip the humans involved. It feels like I'm reviewing the output of their AI assistants on their behalf.

What do I actually do here?"

JP, Ontario

Dear JP,

I've been reading up on this subject as it frustrates me also. Just last week I had a conversation with a revenue ops guru who said something that I had to immediately write down:

The burden of content is now on the reader rather than the creator.

I really felt that was a great way of describing it, and your experience aligns perfectly. This problem is very, very real.

Researchers at Stanford and BetterUp have a name for this now, "workslop," AI-generated work that looks finished but doesn't actually advance anything. Their study, published in Harvard Business Review off the back of a thousand-plus workers and two years of data across 400,000 employees, found that roughly four in ten people have received it in the last month alone, and each instance costs the recipient close to two hours of unpicking, checking or redoing. Scaled across a workforce of ten thousand, HBR put the productivity hit at over nine million dollars a year. You've got three people doing it to you at once. I don't think you're imagining it, at all.

Producing a deck used to be the expensive part. The thinking happened during the writing, because writing forced conclusions, ideas and recommendations to the surface. Now the drafting is close to free, so the thinking that used to happen along the way isn't happening by default. Some workers are simply using AI assistants to replace their thinking rather than improve it.

You're not reviewing your colleagues' work anymore. You're reviewing Copilot's work, on their behalf, without them having done the first pass of thinking that used to happen before something reached you. The irony is that the total time taken overall is probably the same as it used to be, it's just moved to your side of the desk.

Kate Niederhoffer at BetterUp and Jeff Hancock at Stanford, the pair behind the original workslop research, argue in a follow-up HBR piece that management tends to manufacture the problem it then complains about: a vague company narrative to "embrace AI," a named tool everyone's told to adopt, and no real guidance on what good use of it actually looks like versus a shortcut.

That's your company, by your own description. This pattern, and its impact, will compound.

The executive search firm Russell Reynolds Associates' party line is that leadership skill, not AI policy, is the actual bottleneck. In their view, the leaders who stop workslop from spreading are the ones who've learned to recognise when AI is the wrong tool for a task and can coach people on blending its output with real judgement, the thing the company is actually paying that person for.

From the people I have debated this topic with, the consensus seems to be that we have to engage those responsible for producing the reports in good old-fashioned conversation. They have produced the dissertation, now it's time for the viva voce. You're busy, they, apparently, wrote it. Ask them questions on a few key areas and you'll be able to quickly judge whether there is in fact a person behind it or whether it was just a prompt. One COO told me "workslop is hollow at the point of scrutiny when it's polished at the point of delivery."

For your case, it might be worth working with HR to address it. What training would not only get the message across but make the, AI enabled, output much, much better? Before reaching for AI, encourage the team to ask whether it's the right way to do this particular task; that if they can't explain or defend what it produced, not to send it; and when it matters, say plainly how AI was used and what was checked.

Right now, your team sounds like it has the opposite incentive, use AI, don't mention it, hope it 'will do'.

Sod that.

And with all the AI taking jobs narrative, you'll actually be doing them a massive favour in helping them get their relationship with AI assistants right.

Onwards!

Rich