AI work attribution

A record of who actually did the work.

Your timesheet records the person the hours were booked under. It cannot say whether that person wrote the deliverable or an agent did. WhoWorked keeps the whole chain: contributors, agent sessions, models, cost, who reviewed it and what shipped.

  • No screenshots or keystroke logging
  • Provenance per deliverable
  • Exports for a client or an audit

Who worked on this?

Alan R.Drafted38%
Claude OpusGenerated54%
Priya K.Prepared8%
DeliverableQ3 valuation model1.9× leverage

Reviewed by Alan R. before it shipped

The chain behind one deliverable, recorded as it happened

What the entry leaves out

Six hours booked to a person is not a record of who did the work.

A conventional entry names the person the hours sat under. Whether an agent produced the output behind that booking, which model ran, and who checked it before it shipped are all outside what the format can hold.

That gap is fine until a client, a prime or an internal review asks the question directly.

Where provenance disappears

Three questions your delivery record cannot currently answer.

Person or agent?

The booking records a name. It does not distinguish between work someone produced and work they delegated to an agent and then approved.

Reviewed by whom?

Agent output that nobody checked and agent output a senior reviewed look identical once they reach a client. Only one of them should.

Provable how?

MSAs now ask suppliers to disclose the AI contribution to a deliverable. Signing that means attesting to a share most firms reconstruct from recollection.

What the work record holds

Trace a deliverable back through everyone and everything that touched it.

Each deliverable carries its contributors, the agent sessions that ran beneath them, the models and tokens involved, the review that happened before it shipped, and the artifact that resulted. The person stays the headline and the agent work is the detail underneath.

  • Keep contributors, agent sessions and review on one deliverable.
  • Measure the assisted share while the work happens.
  • Export provenance for a client, a prime or an internal audit.
  • Answer disclosure clauses from evidence rather than recollection.

Start with one deliverable and see the whole chain.

Connect the agents working on a single project and the provenance builds itself as the work happens. 30 days, no card, and nothing else in your stack has to move.

How provenance gets recorded.

01

Work reports itself

People log hours the usual way. Agents report their sessions over MCP or REST as they run, so both halves arrive without a second system to reconcile.

02

Contributions attach to the deliverable

Every session and entry joins the thing it produced, which is what turns a list of activity into a record you can actually trace.

03

The record answers the question

When a client asks what share of the work was AI-assisted, the number comes from the delivery record. Pro exports it in a form you can hand over.

FAQ

Start counting all the work.

30-day free trial. Bring the time history you already have.