A number you can argue with.
Every career in the catalogue carries a 0 to 100 rating for how much of its work a machine can already do. This page is how that number is built, so you can decide whether to believe it.
Measured by task, not by job title.
A job title is a bundle of tasks, and automation does not take bundles, it takes tasks. Every career record splits its work into what a machine can already do, what it helps with, and what it cannot touch, and the rating is what falls out of that split. That is why the lists are published next to the number: the number is a summary of them, not a separate opinion.
1,516 rated careers, averaging 43. The right hand tail is thin: the catalogue runs out well before the scale does.
What it actually does.
The task split, in the open
Three shares, and the named tasks under each: what a machine already does, what it assists with, and what it cannot touch. The tasks come from the role's own record, not its job title, so you can see the exact line you disagree with.
Six drivers, each one shown
Digital surface, routine intensity and rule bound thinking push a rating up. People, accountability and physical presence pull it down. All six are printed, so a surprising number traces to the driver that made it.
Read against the Kenyan market
Exposure is what the technology can do. Whether an employer here has bought it is a separate question, so hiring, wages and local adoption sit beside the rating rather than inside it.
Bands, because precision would be false
Low below 40, moderate to 65, high above. Plan against the band: the figure is a composite of six judged scores, so a 62 is not meaningfully different from a 58.
Placed in the distribution
58 reads as frightening until you see the catalogue averages 43. Every record plots its score against all rated careers and against its own sector's mean, because the question is how unusual it is.
How to actually read one.
The rating is a starting point for a decision, not the decision.
- 01
Read the band first
Low, moderate and high are what a plan should actually respond to.
- 02
Check the split
Ask honestly how much of your week the automatable share describes.
- 03
Then the human share
That is the part worth deepening, and usually where the pay went.
- 04
Read the market notes
A capability that exists is not the same as one your employer has bought.
- 05
Follow the moves
Which is where the transition routes and their skill overlap come from.
Who gets it.
Every tool is on every tier. What a tier buys is credits, which is how much model work you can spend in a week.
The catalogue is open today and needs no account. The app itself is in invite-only beta.
Compare tiers- FreeKES 0
- IncludedIncluded
- ProKES 500 a month
- IncludedIncluded
- ProfessionalKES 900 a month
- IncludedIncluded
What it does not claim.
The rating is useful because its boundaries are stated. These are the things it is not.
5 stated
- 01It is not a prediction that a job will disappear. It measures how much of the current task list is automatable, and roles usually change shape long before they end.
- 02It is not a measurement. It is derived from published research applied to recorded tasks, and the newest study behind it is from 2023.
- 03It is not Kenya specific at source. The underlying research is global; the market notes on each record are where local adoption is accounted for.
- 04It is not a live index. Ratings move when the catalogue is reviewed, not continuously, and the record shows what it was built from rather than a daily figure.
- 05It cannot see your employer. Two people in the same role at different organisations face very different timelines, and nothing here knows which one you are.
Common questions.
If yours is not here, ask. We answer in writing so the answer is quotable, and we correct the record when we get one wrong.
Ask usRead the ratings for yourself.
The whole index is open, sorted either way, with the reasoning attached to every entry.