# Enterprise AI Agent 90-Day Roadmap: Board-Ready Template

Prepared for: [organization]  
Executive sponsor: [name / role]  
Workflow owner: [role]  
Technical owner: [role]  
Reporting period: [start date] to [day 90]

## Executive decision

- Workflow selected: [one bounded workflow]
- Business outcome: [cycle time / quality / cost / revenue target]
- Highest-impact action: [the action that cannot run without review]
- Approval owner: [named role]
- Go / hold / stop decision required on day 90: [decision]

## Baseline

| Metric | Current baseline | Day-30 target | Day-60 target | Day-90 target |
| --- | ---: | ---: | ---: | ---: |
| Weekly workflow volume |  |  |  |  |
| Median cycle time |  |  |  |  |
| Quality / error rate |  |  |  |  |
| Cost per completed case |  |  |  |  |
| Approval P50 / P95 latency |  |  |  |  |
| Rejection rate |  |  |  |  |
| Timeout rate |  |  |  |  |
| Verified callback success |  |  |  |  |

## Days 1-30: prove the governance model

- [ ] Select one high-frequency, bounded, measurable workflow.
- [ ] Name the business owner and technical owner.
- [ ] Inventory the agent, model, tools, data sources, and environment.
- [ ] Define autonomous, gated, and blocked action classes.
- [ ] Put the highest-impact action behind routed human approval.
- [ ] Set an SLA, fallback owner, and fail-closed timeout behavior.
- [ ] Verify the signed decision before the agent resumes.
- [ ] Record the request, context, reviewer, decision, callback, and outcome.
- [ ] Start weekly reporting for adoption, latency, rejection, timeout, and callback success.

Day-30 evidence: [links to policy, control map, sample evidence, and dashboard]

## Days 31-60: harden the workflow

- [ ] Review every rejection cluster and update policy where needed.
- [ ] Add audit-only records for authorized autonomous actions.
- [ ] Test missed-SLA escalation and fallback routing.
- [ ] Confirm least-privilege access to tools and data.
- [ ] Run an incident tabletop with business, engineering, security, and legal/governance.
- [ ] Maintain two consecutive weeks of stable operating metrics.

Day-60 evidence: [metric trend, incident findings, policy changes, owner sign-off]

## Days 61-90: decide whether to scale

- [ ] Reuse the same owner / policy / approval / evidence template on one adjacent workflow.
- [ ] Compare day-90 results to the baseline.
- [ ] Quantify value only after operating metrics are stable.
- [ ] Document the next three workflows and owners.
- [ ] Present a go / hold / stop recommendation.

## Board summary

1. What is live, and who owns it? [answer]
2. Which actions remain human-controlled? [answer]
3. What evidence shows the controls work? [answer]
4. What measurable value did the workflow create? [answer]
5. What is the next scale decision and its risk? [answer]

Readiness evidence, not legal advice. Legal/governance teams own final system classification.