ImpactTasksEvidence

Task Graph

The Task Graph connects a role to its tasks, then shows where AI may change the work, what remains human-led, and which evidence supports the signal. It is a role intelligence map, not a job-loss prediction or one opaque automation score.

How this works

A role becomes inspectable when every task keeps its evidence path.

The graph starts with a public role, follows one real work activity, shows the AI work-change read, then keeps the supporting evidence and source visible.

Concrete example path

Accountant -> Reconcile accounts -> Match records -> AI work-change signal -> Public task evidence -> Source

Task by task
  1. 1

    Role

    Start with one public occupation page.

  2. 2

    Tasks

    Break the role into the work people actually do.

  3. 3

    Estimate

    Read the task, not the whole role as one claim.

  4. 4

    Evidence

    Attach supporting evidence when a direct match exists.

  5. 5

    Source

    Keep source records visible for inspection.

  6. 6

    Role page result

    Show the role result without hiding the path.

Impact

The top role read asks how much work may change with AI support for preparation, drafting, analysis, review, or information work.

Tasks

The product works task by task because a role contains many kinds of work, and each one can change differently.

Evidence

Task-level evidence can inform a task view only when it is matched directly. Role-level benchmarks stay contextual.

Current coverage

What is publicly inspectable now

These counts are scoped to the 10 role pages currently open in the public preview, not the broader 1,016-role data layer.

10

public roles available

213

tasks in public roles

Nearly all

task evidence coverage in public roles

250

evidence rows attached to public roles

Source principles

Different sources answer different questions

The public role page keeps source-specific signals visible instead of folding everything into one hidden score.

O*NET task backbone

Defines the public occupation and task structure that role pages are built on.

Direct public task evidence

Used when a public AI task signal joins to the same occupation task key.

Observed task-use evidence

Useful as observed-use context, but not treated as a full automation estimate.

Role benchmarks

Shown as occupation context rather than silently assigned to every task.

Role pages

Where the graph appears in the product

Role pages still prioritize fast task scanning. The evidence graph appears inside a selected task so the user can inspect why a signal is shown without losing the task list.

Scan work areas and tasks first.
Open one task to see what changes and what stays human.
Use the evidence graph when you want the source path.