Inputs
The systems and standard your team chooses
- Runbook, source material, and definition of done
- Records from Box, Confluence, ServiceNow
Policy impact assessment
Proposals traced to affected programs
What it does
- Maps a proposed policy change to programs, rules, and affected populations
- Surfaces implementation dependencies and conflicting authority
- Drafts an impact brief with each conclusion cited
How it works
- 1Reads the proposal alongside current policy and program records
- 2Builds the affected-rule and stakeholder map
- 3Routes the draft assessment to policy and legal reviewers
Deployment contract
What goes in, what comes out, and where people review.
Outputs
A finished artifact with its work attached
- Maps a proposed policy change to programs, rules, and affected populations
- Surfaces implementation dependencies and conflicting authority
Approvals
Human control at the steps that matter
Route exceptions and controlled actions to the designated reviewer before anything is committed.
Evals
Quality measured against your rubric
Score completed runs against accepted examples and task-specific criteria. Review failures with the source trace before changing the runbook or model.
Example result
The result is reviewable, not a black box.
A completed run keeps the task, sources, actions, approvals, and eval result together. Open any step to inspect what happened or send it back with a correction.
RUN
Policy impact assessment
- Reads the proposal alongside current policy and program recordsSource trace saved
- Builds the affected-rule and stakeholder mapSource trace saved
- Routes the draft assessment to policy and legal reviewersOutput prepared
Rubric score, sources, and action log travel with the result
Related agents
Deploy in days, not months.
Request demo