Post-launch decision asset
At day 30, review the operating evidence.
The first month should answer whether the scoped agent works safely and usefully in the real workflow. Review outcomes, corrections, escalation quality, cost, ownership, and incidents before expanding channels, adding roles, or requesting a public proof asset.
30-day review agenda
Four evidence groups before the next investment.
Outcomes against baseline
Compare like-for-like operating periods and explain material demand changes.
- Review volume received, attempted, completed, and unresolved
- Compare turnaround, backlog, and human handling effort
- Separate capacity returned from cash savings
- Record evidence sources and any baseline limitations
Quality and customer impact
High throughput is not useful when corrections or repeat contacts increase.
- Sample outputs against approved sources and acceptance criteria
- Review corrections, reopens, duplicates, and misroutes
- Assess complaints, opt-outs, abandonment, and customer feedback
- Identify recurring failure patterns and their owners
Controls and operating ownership
Confirm that the designed guardrails survived real use.
- Review human approvals, escalation time, and unresolved queues
- Check permissions, logs, retention, credentials, and stop control
- List incidents, near misses, provider failures, and recovery actions
- Confirm policy, data, system, and support owners
Cost and next decision
Use actual operating cost and correction effort, not only the original estimate.
- Reconcile model, channel, infrastructure, licence, and support cost
- Include internal review and exception-handling effort
- Decide continue, correct, narrow, pause, expand, or retire
- Write acceptance conditions for the next review period
Buying model comparison
Expansion requires stable evidence—not spare feature ideas.
The first workload should be accepted before another agent or channel compounds its weaknesses.
Outcome
Ready to consider expansionAccepted KPI trend with explainable variance.
Correct or pause firstBaseline is missing or outcomes are disputed.
Quality
Ready to consider expansionCorrection and failure rates are within accepted thresholds.
Correct or pause firstRecurring errors or unresolved customer impact remain.
Control
Ready to consider expansionApprovals, escalations, logs, and stop path operate as designed.
Correct or pause firstExceptions are bypassed or ownership is unclear.
Cost
Ready to consider expansionActual total operating cost supports the business case.
Correct or pause firstProvider, support, or human review cost is unknown.
Proof readiness
Earn the right to expand or publish.
Operating evidence and explicit permission come before marketing value.
Evidence is ready
- The client has accepted measured outcomes
- Limitations and comparison periods are documented
- Confidential data can be safely removed
- Written proof or referral permission is recorded
Do not publish or expand
- Metrics are estimated or selectively reported
- The observation period is not representative
- Client approval or evidence ownership is unclear
- Unresolved quality or incident issues remain
Buyer objections
Questions to resolve before signing.
Should the review happen exactly on day 30?
Thirty days is a practical first checkpoint, not a universal rule. Low-volume or seasonal workflows may need a longer observation period; high-risk issues should be reviewed immediately.
When can we publish a case study?
Only after the client verifies the facts, the evidence period is representative, confidential details are removed, outcome language is accurate, and explicit publication permission is recorded.
When should we ask for a referral?
Only after the client has accepted a useful outcome and the request is appropriate. A referral should never be tied to hiding limitations or overstating results.
What if results are below the audit estimate?
Reconcile actual volume, automatable share, human review, failures, and recurring costs. Correct the workflow or business case; the audit estimate should not override observed evidence.
Continue the evaluation
Continue your deployment evaluation.
No-pressure scope check
Assess the next workload with the same evidence discipline.
If the first role is stable, use the audit to compare another workload. Expansion should still earn its own operating case.
Start the Agent ROI Audit