Research estimateWork activitiesEvidence

From a role title to work you can inspect

The Task Graph turns a role into recognizable work activities, connects each one to public AI research, and keeps real workflow context separate. Use it to understand what deserves closer investigation—not as an automation instruction or job forecast.

How this works

Connect the role, the work activities, and the evidence behind each one.

Start with a public role, follow one recognizable activity into its technical-exposure class and other evidence, then add the real work context research cannot know.

Illustrative evidence path

Accountant → Reconcile accounts → OpenAI E0/E1/E2 class → Other evidence → Real workflow review

Task by task
  1. 1

    Public role

    Start with a public role or the closest recognized match.

  2. 2

    Work activities

    Open the recurring work inside the job title.

  3. 3

    Technical exposure

    See whether the threshold is met directly, needs AI tools, or was not met.

  4. 4

    Other evidence

    Add direct task matches without blending their meaning.

  5. 5

    Real context

    Add context supplied by the person or organization; never infer it from a public baseline.

  6. 6

    Next question

    Use the evidence to focus a decision; do not treat it as the decision.

Roles are starting points

A public role provides a consistent baseline. It does not reveal how one person or organization performs the work today.

Tasks reveal the differences

The product works activity by activity because one role contains many kinds of work, and each one can carry different evidence and constraints.

Evidence stays separate

OpenAI task classifications form the headline. Anthropic observed use, EconEvals modeled time saving, O*NET context, and role benchmarks stay separate.

One foundation, three uses

Start with a public role. Add real work only when the person or organization provides it.

Each use starts from the same task-level research and visible evidence, but supports a different question and keeps personal and organization data separate.

Public role pages

Open, inspectable baselines for seeing the work inside a role and identifying activities worth closer investigation.

Explore roles

Personal analysis pilot

Invitation-code reports for mapping your recurring work, approving public matches, and exploring workflow or capability questions.

Map my work

Planned organization product

A separate product for mapping organization-provided roles, tasks, and complete workload data to focus workflow, oversight, and capability questions. It is not available yet.

For organizations

Current coverage

What is publicly inspectable now

These figures cover the 10 role analyses currently available, not every role in the broader 1,016-role catalog.

10

public roles available

213

tasks in public roles

100.0%

task evidence coverage in public roles

633

evidence rows attached to public roles

Go deeper

The overview shows how the analysis fits together. The methodology explains the calculation.

Read the methodology for the beta technical-exposure score, evidence rules, and limitations. Read the Sources page for dataset versions, matching methods, and source-specific caveats.

Role pages

Where the graph appears in the product

Role pages prioritize fast task scanning. Open a task to inspect its study result, exact match, other findings, and source links without losing the task list.

Scan work areas and tasks first.
Open one task to see its modeled time-saving band.
Check the match, other findings, and source links when you need detail.

Your task graph

Want to compare your own tasks with the research?

Describe your recurring work, confirm the closest public role, and review every task match before the analysis is generated.

Map my work