Your profession is not one thing — it's a bundle of tasks. A marketing manager drafts content, reports on campaigns, sets strategy, manages relationships. We measure AI exposure at that task level, not as a single verdict on the job.

Here is the join, start to finish. We begin with the task statements from O*NET, the U.S. Department of Labor's occupational database — the standard breakdown of what each occupation actually does, keyed to O*NET-SOC occupation codes. To each task we attach usage data from the Anthropic Economic Index, which measures how AI is actually being used across work, and whether that use looks like automation (AI doing the task) or augmentation (AI assisting a person doing it). We add wage and outlook figures from the Bureau of Labor Statistics — median pay, projected growth, typical openings — rolled up from the O*NET-SOC codes to the broader SOC codes the BLS reports on.

All of this is precomputed into static data files ahead of time. When you look up your role, you're reading numbers that were already assembled from these public sources — nothing is invented on the spot. Where the site uses AI to narrate your report, Claude describes what the numbers say; it never generates the numbers. That line is not blurry: the data is the data, and the writing is labeled as writing.

From the task-level exposure, every task sorts into one of three groups. This framing is the heart of the report:

The single most important rule on this site: exposure means observed AI adoption in the tasks your role involves — it is not a probability that your job disappears. Those are different things, and conflating them is the defining mistake of the genre.

Task-level AI use is something we can measure from real data. Whether a specific person loses a specific job depends on their employer, their team, the economy, their own choices, and a hundred things no dataset captures. So we do not publish a single scary percentage of "your job is X% at risk." Tools that do — the ones built around one alarming number — are the anti-pattern we're deliberately avoiding.

What you get instead is a truthful, task-level picture of where AI is showing up in work like yours, sorted into what's automating, transforming, and durable — so you can act on the parts you can influence.

The data lags reality. AI adoption is moving faster than any dataset can be published. The sources here are the best public measures available, but they are snapshots — by the time you read a figure, the ground has moved somewhat. We date everything so you can judge how fresh it is.

Task-level is not your specific job. Two people with the same job title do different work, on different teams, with different tools. The report describes the typical task mix for an occupation, not your particular role. Read it as a map of the terrain, not a readout of your exact position on it.

Coverage is partial at launch. We start with a focused set of white-collar and knowledge occupations — roughly two hundred where the underlying AI-usage data is densest — rather than every job in the economy. If your exact role isn't covered yet, the closest match will still be informative, and coverage grows over time.

None of this makes the picture useless. It makes it honest — which is the only kind worth acting on.

Some surfaces on this site use Claude, Anthropic's AI, to help you make sense of your data. They are:

The Signal narrative — the written summary that explains your task-level report in plain language.
Tool coaching — the prompts and feedback inside the Pivot tools (skills, values, energy, stories).
The Career Lab — longer-form drafts you can generate and then edit.

Two rules hold everywhere AI appears. First, AI output is labeled as AI-generated wherever it shows up — you always know when you're reading a machine's words rather than a cited fact. Second, it narrates, it doesn't invent. The exposure figures, wages, and outlook come from the sources above; Claude describes them but never makes up a number. Everything it writes is a starting draft you can edit and own.

For the full account of who's behind the project and the promises around your privacy, see the About page.