The core idea: tasks, not jobs
A job is a bundle of tasks. AI rarely takes a whole job; it takes some tasks and leaves others. So instead of scoring whole occupations (which tends to produce scary, misleading numbers), this site scores the 68 tasks people actually do, and reads each field as the mix of tasks its work is built on.
The four bands
Every task and field sits in one of four bands. It is a heat scale of how exposed the work is to current AI, not a judgment of its worth:
- Machine territory AI already does this well for most everyday cases. The task itself is largely automatable.
- Moving fast AI is getting capable here quickly. This is where the pressure on the craft is climbing fastest.
- Shared ground AI assists strongly, but a person stays in the loop, accountable for the judgment call.
- Human ground Rooted in presence, trust, the body or accountability. The least exposed work on the board.
Field bands use the same colours under different names: High exposure, Rising exposure, Mixed and Grounded.
What the bands are, and are not
The band for each task and field is an editorial synthesis of the research below, not a score from any single study. It reflects a considered read of where current AI is capable, weighted toward what the evidence broadly agrees on. It is a starting point for a conversation, not a measurement. Reasonable people will move some cards a band over, and that disagreement is exactly what the cards are for.
Exposure is not the same as replacement. A high band means more of the work is technically automatable, not that the job disappears. See the signal Exposure is not the same as replacement.
The seven signals
The load-bearing findings behind the bands, each with its evidence and an honest counterpoint:
The study shelf
The research these pages draw on. Every figure was checked against its primary source on June 12, 2026; titles, years and URLs point at the canonical versions.
Estimated 47% of US jobs were at high risk of automation over the following one to two decades, scoring 702 occupations.
Using a task-based approach, estimated only about 9% of jobs (not 47%) are at high risk of automation, because most jobs mix automatable and non-automatable tasks.
About 14% of jobs across OECD countries are highly automatable, and a further 32% face significant change to how they are done.
Around 80% of the US workforce could have at least 10% of their work tasks affected by LLMs, and roughly 19% could see at least 50% of tasks affected. Higher-wage, knowledge work is more exposed than physical work.
From millions of real Claude conversations: roughly 57% of usage was augmentative (working with the user) versus about 43% automative (doing the task), with use concentrated in software and writing tasks.
Employers expect 170 million new roles created and 92 million displaced by 2030 (a net 78 million gain). Analytical thinking remains the most sought-after core skill, while AI and big data, networks and cybersecurity, and technological literacy are the fastest-growing skills, with resilience, flexibility and agility also rising.
Current generative AI and other technologies combined could automate work activities that absorb 60 to 70 percent of employees' time today, with about 75 percent of generative AI's value concentrated in customer operations, marketing and sales, software engineering, and R&D.
Generative AI could expose the equivalent of 300 million full-time jobs to automation, while also lifting productivity and creating new work.
Automation substitutes for some tasks but complements others and raises demand for the labor it cannot replace; jobs are bundles of tasks, so automating a task rarely deletes the job.
About 60% of employment in 2018 was in job titles that did not exist in 1940; new work is continually created, much of it driven by new technology.
Since 1980, jobs requiring high social skills grew strongly and paid more; tasks needing teamwork and interpersonal interaction have been hardest to automate.
A generative AI assistant raised customer-support agent productivity by about 15% on average, with the largest gains, around 30%, for the least experienced and lowest-skilled workers.
Most jobs are more likely to be augmented than automated; clerical work is the most exposed category, and effects fall unevenly across regions and genders.
Built on MethodKit
The fields, tasks, skills and prompts are the cards of four MethodKit decks: Topics (fields), Competencies (tasks), Future Skills (safer ground) and Professional Development (the self-audit). The decks are discussion tools; this site arranges them around one question. Browse the kits →