SCOPED PILOT SERVICES FOR AGENCIES
Check the action.
Explain the decision.
Changing an agent or investigating an unexpected action? Start with one workflow: check the business rules, inspect the evidence and prepare a clear client handover.
Pilot enquiries are open. The software platform is in early development. We confirm delivery capability, integrations and scope before accepting work.
Discuss your workflow ↗Choose the support your release needs.
01Check a release
Changing a prompt, model or tool? Compare the change against the approvals and business rules your client relies on.
- Test setup: one workflow, agreed rules and a baseline
- Approval, record accuracy and failure-recovery checks where supported
- Release review: evidence, open issues and a human release decision
A scoped comparison and clear release findings.02Investigate a failure
Bring an unexpected agent action. Assess whether it can be reproduced and turned into a check for future releases.
- Review the reported issue, relevant traces and starting state
- Attempt reproduction in an authorised test environment; record any limits
- Agree a repeatable check where feasible; fixes and reruns scoped separately
A documented finding and an agreed recheck plan.03Maintain the checks
Keep useful checks as client workflows evolve, with reports that make the next handover easier.
- Ongoing coverage: agreed runs and test maintenance allowance
- Add confirmed failures to the suite within the agreed scope
- Agency-branded reports where agreed, with findings, owners and limits
A proposed plan for ongoing coverage and handover. BEYOND THE HAPPY PATH
What happens when a step fails?
A useful check looks beyond a reassuring answer. Depending on the agreed environment, we can investigate:
- Refunds or messages attempted before matching approval.
- Updates targeting the wrong customer or stale record.
- Timeouts followed by duplicate payments, bookings or messages.
- Failed actions reported as complete instead of escalated.
Read the failure-recovery checklist →Your agency. A clearer handover.
Discuss agency-branded reports with the checks, findings, open issues and next actions your client needs. Branding and editable deliverables are agreed in scope.
Build on the tools you use.
Keep the agreed business rules and findings available to your team. We assess compatibility with your existing test setup before proposing changes.
Read the sample report and get the checklist →THE DELIVERABLE
A release review you can use at handover.
The pilot scope sets out what evidence can be collected. A proposed report includes:
- The business rules and agent versions being checked.
- Scenario results, including failures, skips and incomplete runs.
- Supporting tool evidence and observed state changes where the setup allows.
- Unresolved issues, recommended next steps and a record of human review.
Where only exported traces are available, we describe the work as trace analysis. It does not prove a live workflow completed or an external record changed correctly.
Explore the report concepts →Fixes and reruns
Need help resolving a finding? We can discuss a separate engineering scope and rerun. Access, permitted changes and acceptance are agreed first.
Your tools first
We review your current setup before proposing new tools. Compatibility and third-party costs are confirmed during scoping.
A passing test has limits
A report covers the scenarios and conditions tested. Your team retains the release decision and appropriate human approvals.
BEFORE ANY WORK STARTS
Know the scope.
Know the commitment.
- One workflow, the test cases and the expected deliverables.
- A suitable test environment, data access, redaction and retention arrangements.
- The fee, run allowance, third-party costs and delivery timing.
- What is excluded and how additional work is approved.
Start with a scoped setup or release review. Ongoing coverage is discussed when your release frequency and maintenance needs justify it. There are no published subscription plans or fixed turnaround guarantees.
Bring one important scenario.
Tell us what the agent does, what is changing and the action it must get right.
Discuss a pilot ↗Explore the illustrative report →