◆ AGI ScorecardExposure check 📋
Will AI take my job?

Check your AI exposure in 60 seconds

Six questions about how you actually work. You get an honest exposure tier — grounded in what 2026 systems really do (~83% on knowledge work, ~80% on agentic coding) and the one line they haven't crossed: the autonomy gap. No sign-up, no invented percentages.

Question 1 of 6
Your AI exposure tier
✓ Copied

Your exposure becomes replacement only when AI crosses the autonomy gap — and we track that line, live.

Get told when the line moves →
📋 AI Job Risk Check · See how the tiers are scored on AGI Scorecard →

How the tiers work

🟢 Sheltered (for now) — your work sits behind the strongest shields: physical presence, human trust, and end-to-end accountability.
🟡 Task-exposed — parts of your job overlap what AI already does well, but the core still needs a human in the loop.
🟠 Benchmark-adjacent — much of your work looks like what's measured on GDPval (~83%) and SWE-Bench Pro (~80%). The autonomy gap is your main shield.
🔴 In the crosshairs — screen-based, structured, reviewable work with little trust or judgment moat. Exposure is already here; replacement waits on the autonomy gap.

The rubric is transparent: each question maps to evidence tracked on this scorecard — benchmark performance, the robotics lag, the human-trust moat, and the undemonstrated autonomy milestone (AGI-2027 Thesis Tracker: 62.5/100). It scores your answers; it doesn't invent per-job statistics. Deeper reads: can AI replace knowledge workers? · will AI replace programmers? · mass unemployment?

FAQ

How does the AI job exposure check work?
Six questions score the parts of work that 2026 evidence says matter: how screen-based and structured your tasks are (AI already reaches ~83% on GDPval knowledge-work tasks and ~80% on SWE-Bench Pro agentic coding), versus the shields — physical presence, human trust, and unsupervised end-to-end judgment, which no system has reliably crossed. You get an exposure tier, not an invented percentage.
What is the autonomy gap?
The gap between scoring well on benchmarks and reliably owning a whole job end-to-end without supervision. As of mid-2026 no AI system has crossed it — it is the single line that decides when task exposure becomes actual job replacement, and it is what the AGI Scorecard tracks.
Is this based on real data?
The rubric is transparent and grounded in documented evidence tracked on this site: ~83% on GDPval-style knowledge work, ~80% on SWE-Bench Pro, and the undemonstrated autonomy milestone (AGI-2027 Thesis Tracker at 62.5/100). It returns a tier based on your own answers — it does not invent per-job statistics.

Put this check on your site

Free to embed on any blog, newsletter or careers page. It keeps the six questions and the tier result, drops the site chrome, and carries one link back. More tools at /ai-tools.

✓ Copied