Application availability
99.92%
Target ≥ 99.90%
- Source:
- Synthetic uptime-probe events
- Cadence:
- Five-minute checks
- Owner:
- Platform operator
- Monitoring Mechanism
- Failed probe → incident runbook and human acknowledgement
The Evidence Room
Not activity theatre. Inspectable evidence.
Every programme mission produces an evidence pack: baseline, target, result, source, Monitoring Mechanism, owner, limitations and client acceptance. Explore the format without mistaking demonstration data for client results.
Sample Mission Evidence Pack
A sample pack for one programme mission.
This pack demonstrates format only. It is not a case study, live result or promised performance level.
| Mission | Starfighter — private AI pilot | Month 06 / BASTION |
|---|---|---|
| Business outcome | Draft support triage with human approval | One bounded workflow |
| Baseline | Median handling time: 11 minutes | 40 synthetic sample tickets |
| Target | Median ≤ 8 minutes; quality ≥ 90/100 | Agreed before the pilot |
| Result | 7.5 minutes; quality 92/100 | Synthetic demonstration result |
| Data source | Workflow event log + human evaluation set | Source and version recorded |
| Monitoring Mechanism | Daily exceptions; weekly sampled evaluation | Named Applied-AI operator |
| Limitations | Synthetic data; no production or client claim | Real packs state their own limits |
| Client acceptance | Accepted / revise / stop | Decision and exceptions signed off |
Quarterly Mission Control
Show the mechanism behind the number.
Source, cadence, owner and action travel with the KPI. A missed target stays visible; it is not edited into a success story.
Demonstration only — every figure below is synthetic, not live client telemetry or a Cooper Agile performance claim.
Automation success
96.4%
Target ≥ 97.0%
- Source:
- Synthetic workflow event log
- Cadence:
- Daily aggregation
- Owner:
- Applied-AI operator
- Monitoring Mechanism
- Exception queue → sampled review and corrective FORT
AI task quality
92 / 100
Target ≥ 90
- Source:
- Synthetic human evaluation set
- Cadence:
- Weekly sampled evaluation
- Owner:
- AI solutions lead
- Monitoring Mechanism
- Quality regression → model/version review before promotion
Backup success
100%
Target = 100%
- Source:
- Synthetic backup job records
- Cadence:
- After every scheduled job
- Owner:
- Service coordinator
- Monitoring Mechanism
- Failed job → retry, escalation and recovery-test schedule
Quarterly decision record
Scale, revise or stop. On evidence.
Real reviews include targets missed and changes made. The evidence exists to support a decision, not to defend a retainer.
Scale
Quality and operating value clear the agreed threshold. Increase scope deliberately, with cost and risk still visible.
Revise
The use case remains promising but data, workflow or human-review controls need another bounded FORT.
Stop
Value, adoption or risk does not justify continuation. Close the experiment, document the learning and release the capacity.
Configure a programme with this evidence discipline built in