Free Tool — The AI Label Test
Is it actually AI?
One flagship project, twelve questions, two axes: does the system genuinely make AI-shaped calls, and does it have a value case either way? The verdict can't be charmed — and a few answers raise flags all by themselves. Runs in your browser; nothing you type leaves it.
12 questions. ~6 minutes. No email, ever. Finished verdicts are stored anonymously — selections only, never the project name — to publish the patterns.
Question 01 / 12
Question title
Question text
Score the proposal, not the deck.
The verdict on
Verdict
The two axes
Take the verdict to the committee
One page, meeting-ready: the verdict, both axis scores, the raised flags, and the next moves. Print it or save it as a PDF — this is the page per project a selection review runs on.
Generated in your browser. The project name never leaves it — not even for this.
Passed the label test? The next trial is delivery — run the pilot through the free Ship / Fix / Kill triage
Put the verdict where your team plans
Copy a one-line snippet with the verdict — Markdown for wikis like Confluence or Notion, plain HTML for a site or intranet page. A link in a project doc is how the next team finds this.
The snippet carries only the verdict and a link. The project name and your answers stay in your browser.
About this test
What the label test checks
A significant share of enterprise “AI projects” are not AI. They are BI dashboards, data-warehouse builds, or workflow automation wearing an AI badge, selected by committees whose incentives reward the label. This test runs one flagship project through twelve questions on two axes:
- Is it actually AI? Does the system make AI-shaped calls — probabilistic outputs, learned behavior, decisions under uncertainty — or does it follow rules a developer wrote?
- Is it a good project either way? A value case, an owner, and a metric — or a slide?
The three verdicts
Crossing the two axes gives one of three verdicts: an AI project (it is AI, and it has a case), good software wearing the wrong badge (worth doing — just stop calling it AI, because the label sets expectations the project can't meet), or a keyword project (it is on the list because someone said “AI”). The second verdict is the one committees least expect and most need: the label, not the project, is often what fails.
FAQ
Do I need to give an email address?
No. The verdict is shown immediately.
Why does mislabeling matter if the project is useful?
Because the famous AI failure statistics are partly measuring projects that were never AI to begin with — and because an “AI” label buys a project scrutiny, budget expectations, and governance obligations that plain good software doesn't need.
The project name I type — where does it go?
Nowhere. It stays in your browser and is never sent; it only makes the verdict read like a verdict.
What should I do with a “keyword project” verdict?
Ask how the project reached the flagship list. The origin questions in the test usually contain the answer — and the fix is in the selection meeting, not the project team.