Personaas

Everyone's using AI. So why does the work feel slower?

A document lands in your inbox looking finished. The analysis doesn't hold, the numbers aren't sourced, and rebuilding it takes you the afternoon. You just received workslop.

A document lands in your inbox from a colleague. It looks great — clean headings, confident prose, a tidy set of recommendations, the right length. You start reading properly, and something curdles. The analysis doesn't actually hold. The numbers aren't sourced. The recommendations could have been written about any company on earth. What looked like a finished piece of thinking turns out to be a shell of one, and now you have to rebuild it from scratch — which takes you the better part of the afternoon.

You just received workslop. And you are very much not alone.

The productivity paradox nobody warned you about

AI was sold as the thing that would make everyone faster. And individually, it does — the person generating the document really did save an hour. The problem is that the hour didn't disappear. It moved. It landed on the three colleagues who now have to figure out whether the confident-sounding output is actually right, and redo it when it isn't.

This is what makes workslop uniquely corrosive. Normally, low-effort work looks low-effort — a half-finished draft announces itself, and you calibrate accordingly. AI breaks that signal. It decouples effort from polish, so a thirty-second prompt can produce something that looks like a day's work. The reader can no longer tell the difference by looking, so they have to check everything. Multiply that across a company and you don't get a productivity boom. You get a verification tax that quietly eats the gains.

AI didn't remove the work. It disguised the work, and mailed it to the person sitting next to you.

There's a slower, more expensive second-order cost too. The HBR research points to it directly: as slop circulates, it degrades the organisation's shared knowledge. Decisions get made on analysis that was never real. Trust between colleagues erodes — you start assuming anything you're handed might be hollow, so you re-verify by reflex. The company gets measurably less sure of what it actually knows. For a business whose edge is supposed to be its intelligence, that's the worst possible thing to be leaking.

Why generic AI produces slop by default

Slop is what you get when AI has three things missing at once:

Add more of that kind of AI and you don't cure workslop — you manufacture it at scale.

Grounded, accountable AI is the actual antidote

The fix isn't less AI, and it certainly isn't "ban the tools" — that just pushes the slop underground. The fix is AI that's built to produce substance instead of the appearance of it. That's the whole distinction between a chatbot and a Personaas persona, and it maps almost exactly onto the three things above.

More from the blog