How it works
A job is a list of responsibilities. We look at each one.
Some parts of a job are things AI already does well. Others need a person. HumanLens splits the job apart, checks every piece, and adds it back up.
Every responsibility, sortedBreak the job into responsibilities
Every job is matched to the U.S. Department of Labor's official list of responsibilities for that type of work. An executive assistant role has 22 responsibilities, from managing the calendar to greeting visitors. If you paste a posting, the responsibilities it stresses count for more.
Ask 4 questions about each responsibility
The questions are below. Every answer comes with its reasons.
Work out how much of each responsibility AI could take on
Few responsibilities are all-or-nothing. Each one gets a share AI could take on, and one of 3 outcomes: AI can do it (it can take on most of the responsibility and gets it right), AI helps (a person still does it, faster, with AI), or needs a person.
Add it up
You see how much of the job's time sits on each side, and what the AI side really costs once you count the hours someone spends checking its work.
The 4 questions
Can AI do it?
Does today's AI produce this responsibility's output to the standard a competent worker would?
Does it get it right?
How often is that output correct, and how easy is a mistake to spot?
Must a person sign for it?
Does the law, a licence, a signature or being physically there require a person?
Does someone need a person?
Is there a customer, client or colleague who needs to deal with a human in the moment?
2 responsibilities from the same job
Both belong to an executive assistant. They end up on opposite sides.
What you get
- For a job: the share of the job's time that stays human, the responsibilities on each side, and the yearly cost of handing the AI side to software, with checking time included. For a typical executive assistant, AI could take on about 60% of the time, and 40% still needs a person.
- For a person: a Human Edge score from 0 to 100. It measures how much of your experience is in the responsibilities that stay human for that job. It looks only at the responsibilities on your résumé, never at who you are.
Good to know
- This is an early preview, and new jobs are added every week based on what employers are posting. The scores are being checked against the judgment of working recruiters, and they'll shift as that feedback comes in. Where a responsibility could go either way, the result marks it as a close call.
- AI gets better every month, so the scores are updated on a schedule. Every result shows which version it used.
- HumanLens helps people decide. It never rejects, ranks or picks anyone by itself.
A living index
The scores move as AI gets stronger.
A score from last year is already out of date. Every month we update the scores from 3 kinds of evidence, and every change records where it came from.
- What people really use AI for. When Anthropic publishes new data showing how often AI succeeds at a responsibility in real work, its score moves with it. Real use beats predictions.
- How fast AI is improving. METR measures how long a piece of work AI can complete on its own, and that length has been doubling about every 4 months. Responsibilities with no new direct evidence move with that trend, slowed right down, because real work changes far more slowly than benchmarks. We measure how much slower from real-world data: AI's success rate on real work rose only about 1.4 points in 3 months, so only about a tenth of benchmark gains count.
- What recruiters tell us. When people who hire for a role keep disagreeing with a responsibility's score, it goes to a person for review. It never changes automatically.
- What never moves on its own. Legal and licensing requirements, and the need for a trusted person in the room, change only when the law or the job does. We update those by hand.
For the detail-minded
The full working
How the scores are put together, and where the data comes from. Every result also explains each of its own scores.
The formulas
Each responsibility is scored 0 to 100 on Capability (C), Reliability (R), Accountability (A) and Trust (T).
Share AI could take on = C/100 x (1 - max(A, T)/100)
Only the strongest reason a person is needed counts. A responsibility that needs a signature doesn't need a person twice as much because a client also likes a human touch.
AI can do it: share of 60% or more and R of 70 or more
Needs a person: share under 25%
AI helps: everything in between
Job share = sum of (weight x share) / sum of weights
Weights come from the Department of Labor's importance and frequency ratings, adjusted to how much your posting stresses each responsibility.
Savings = H x w - (tools + H x (1 - R/100) x w)
H is the yearly hours AI could take on, w is the hourly cost of the role (median wage x 1.3 for benefits and overhead, over 2,080 hours), and R is the reliability of that work. 1 unreliable hour of AI output costs 1 hour of a person's time to check.
Human Edge = 100 x sum(w x (1 - share) x min(1, e / 0.67)) / sum(w x (1 - share))
e is the evidence on your résumé for each responsibility: not shown 0, some exposure 1/3, a regular part of a role 2/3, led or deep 1. Responsibilities count by the part of them that still needs a person. Doing something as a regular part of a role counts as full evidence, and the fairest score comes from checking against the posting you're applying for.
Sources: O*NET 31.0 job data, BLS wages May 2025, Anthropic Economic Index (June 2026 release), and published studies cited per responsibility.
How the scores are checked against research
Our responsibility scores decide which responsibilities AI can take on and why. The overall level is then checked against the best published research, so the numbers match what the industry expects:
- Level: across the first 20 jobs we calibrated, AI could take on 57% of the time on average. That matches McKinsey Global Institute's 2025 estimate that current technology could automate 57% of US work hours.
- Order: where a job's score disagrees with published research on how exposed that job is to AI (Eloundou and others, 2023), we move it halfway toward the research. Our ranking of those jobs agrees closely with theirs, and every new job is checked the same way.
- Possible is not the same as adopted: these numbers show what AI could take on today, not what most companies have done. The U.S. Census Bureau found 19.8% of businesses were using AI as of May 2026.