# AI Workflow Readiness Checklist

Use this checklist before connecting an AI agent to company systems.

## Outcome

- [ ] The workflow has a clear trigger.
- [ ] The completed outcome can be observed in a system of record.
- [ ] The workflow occurs often enough to evaluate.
- [ ] A business owner accepts responsibility for the outcome.

## Process

- [ ] Required inputs and authoritative systems are known.
- [ ] The normal path is documented.
- [ ] Common exceptions and missing-data cases are documented.
- [ ] Escalation owners and response windows are named.

## Permissions and approvals

- [ ] Read and write permissions are separated.
- [ ] Each action is marked permitted, approval-required, or prohibited.
- [ ] Service accounts use least privilege.
- [ ] Sensitive actions identify a named approver.
- [ ] Delegated authority has a defined validity period.

## Evidence

- [ ] Inputs and sources remain traceable.
- [ ] Tool calls, results, retries, and changed records are logged.
- [ ] Approvals record the decision and decision-maker.
- [ ] Final system state is verified.
- [ ] Retention and deletion requirements are documented.

## Pilot

- [ ] A representative evaluation set exists.
- [ ] Expected outcomes were written before testing.
- [ ] Completion, intervention, error, escalation, time, and cost are measured.
- [ ] The pilot has an agreed start, end, and expansion decision.

Source: Dialogo, reviewed 2026-08-15. https://www.dialogo.chat/resources/ai-workflow-readiness-checklist/