Skip to main content

Methodology

How we grade evidence

Short version: we separate what AI can do from what's actually deployed, we grade every claim, and we tell you when the evidence is thin.

Three things, kept apart

Most writing about AI and jobs collapses three different questions into one number. We keep them separate, because they move at completely different speeds:

  • Capability — what an AI system can demonstrably do against a task, measured against a human baseline.
  • Deployment — what share of employers have actually rolled that out. Almost always far lower than capability.
  • Workforce impact — what happened to the people. Slower again, and the hardest of the three to evidence.

The distance between the first two is the useful number. It is why a task can be 90% automatable and still be most of someone’s week.

Every claim carries a grade

Nothing on this site is asserted without a confidence level attached:

  • Confirmed — multiple independent sources, or one peer-reviewed or independent benchmark with published methodology.
  • Plausible — one credible source, or a vendor benchmark that at least shows its method.
  • Early signal — preliminary and unverified. Real enough to watch, not enough to act on.

A source can never be graded above what it can support. A vendor’s own marketing page does not become Confirmed because we would like it to be.

Where we’re weak

Our capability evidence is stronger than our deployment evidence. Most adoption data we hold is early-signal, which means the honest reading of a deployment figure is “barely deployed yet” rather than a precise percentage — so that is how we present it, in words rather than false decimals.

Coverage also varies by role. Where we have published little about a role, the page says so rather than implying a body of work that is not there.

We publish the misses

Predictions are dated, public, and scored against what actually happened — including the ones we got wrong. A track record you cannot check is not a track record. See the calls.

What this is not

This is career intelligence, not career advice. We will tell you what the evidence says about your work. What to do about it stays your decision — we do not know your mortgage, your visa, or how much you like your team.