Workslop: the AI output nobody actually uses

There’s a name floating around the AI world right now: workslop.

It’s what happens when AI generates polished, plausible, professional-looking work that nobody actually USES. Tons of motion. No decisions. Inboxes fuller, calendars heavier, output up… and somehow nothing actually moves.

Sound familiar? Yeah. It’s possible that it’s happening in your business right now…

Monday play

Audit one AI workflow in your business this week. Ask the person running it: “What decision is this output supposed to help someone make?” If they can’t answer in one sentence, the workflow is producing workslop. Either tighten the brief or kill the workflow. Don’t keep running it because it feels productive.

Tool worth a look

Claude Routines (and ChatGPT’s scheduled tasks). Both platforms now run agents on a cron. Daily. Weekly. Triggered by events. Use it for a Friday afternoon “what slipped this week” agent that scans your project tool, email, and Slack and surfaces what’s behind. Most operators waste Sunday nights doing this manually. Stop.

Prompt to steal

Act as a skeptical chief operating officer. I’ll paste a workflow I’m running below. Push back on it. What’s the actual decision someone makes with the output? Is AI the right tool for this step, or am I using AI because it’s trendy? What would you cut, automate, or hand to a human instead?

Drop your workflow in. Let it grill you. Better than a $400/hr consultant.

Mini lesson

Workslop is the AI version of meeting bloat. Tons of motion, no decisions made. The fix is the same fix as meetings: every output needs an owner and an action. If nobody’s going to act on the output, don’t generate it. Faster garbage is still garbage.

Mental model

Input quality > model quality. A weak prompt run through Claude Opus produces worse output than a strong prompt run through GPT-3.5 (that’s an old model btw). People keep upgrading their model when they should be upgrading their brief. The skill that compounds is writing the brief, not picking the tool.

Steal from a podcast

Peter Steinberger (OpenClaw) on AI agent rollouts: agents without strong human direction “still produce garbage, just faster.” His fix is what he calls grilling. Make the AI interview you about the task before it starts working. Anti-vibe-coding. Produces dramatically better results.

The number

$0. That’s roughly the time savings 4 out of 10 workers report from using AI at their job. Not because the tools are bad. Because nobody told them what good looks like.

Heard this week

Some companies are now paying more for AI than they’d pay a human doing the same task. The cost curve is real. The cure isn’t a cheaper model. It’s a sharper thesis.

Talk soon.

Kevin