Is AI making your team worse at their jobs? Ask what they forgot while it worked.

I don't know about you, but this scares me more than the doomsday narratives about AI taking my job. What it means, in effect, is that the brains we rely on to build the thing we're told we'll all have to rely on may already be in decay.

A Lancet study found skill loss in experienced professionals after only a few months of routine AI use. Researchers have a name for the mindset behind it: metacognitive laziness. It's a very articulate way of saying "brainrot".

MIT's "The GenAI Divide: State of AI in Business 2025" report found that 95% of enterprise generative AI pilots fail to deliver measurable returns or scale beyond experimentation.

And headlines about companies pulling back on AI spending because "AI is too expensive" are, I think, a distraction. The real story is what companies are losing while they come to grips with the cash bleed:

What did your people forget how to do while the AI was doing it for them?

A Carnegie Mellon and Microsoft Research study found that the more confidence people place in AI, the less critical thinking they report doing, and that people using AI produce a less diverse set of outcomes. (I didn't need a study to figure that one out. I can just scroll through my LinkedIn feed.)

The erosion of human relational intelligence is the real issue at the heart of AI adoption, and I believe it's a matter of individual responsibility. As with anything on the world stage of ethical debate, it comes down to each of us putting into the world what we'd like to see in it. It comes down to integrity, right down to how we use AI to do a task, serve a client or solve a problem.

What integrity in AI use looks like as a workflow

So what does that look like in practice? Not as a concept, but as a workflow.

I've been working with Anthropic's Claude Cowork Legal Plugin. Legal practitioners are, by professional obligation, the last line of defence between AI output and real-world consequence. They can't afford metacognitive laziness, and neither can their clients. The plugin uses structured AI workflows for contract review, NDA triage, compliance checks and risk assessment, built around human judgement rather than replacing it.

I built a free, practical quick-reference guide to the Claude Legal Plugin on one conviction: AI is only as powerful as the human intelligence directing it.

What this means for your business

If AI is going to do more of the work in your business, capture the judgement behind that work while your people still practise it. The Process Automation X-Ray gets it out of their heads and onto the page, one process at a time.