Deploy Phase Documentation
What Happens in This Phase
The Deploy phase activates software in production and makes it available to users. Documentation includes operational runbooks, incident response procedures, and end-user guidance. AI generates runbooks from operations code while humans ensure user documentation matches actual user mental models and operational needs.
Where AI Helps
- Generating operational runbooks from monitoring configs and scripts
- Creating incident response procedures from alert definitions
- Documenting configuration management from deployed state
- Producing monitoring dashboard documentation from observability setup
- Drafting end-user documentation from feature specifications
Example: Operational Runbook
This example demonstrates GenAI Generation—AI synthesizes alert definitions and remediation scripts into operational procedures that require SRE validation but follow predictable structure.
Context Setup
Before prompting, the SRE provided:
- Prometheus alert definitions (/monitoring/alerts.yml)
- Remediation scripts with inline documentation (/ops/scripts/)
- Escalation policy from on-call rotation
- Service architecture diagram for dependency context
Prompt
"Generate an operational runbook for the mission planning service based on the Prometheus alerts in /monitoring/alerts.yml and the remediation scripts in /ops/scripts/. Include alert descriptions, severity levels, investigation steps, and remediation procedures."
Output
# Mission Planning Service Runbook
## Alert: HighLatencyP99
**Severity:** Warning
**Threshold:** p99 latency > 2000ms for 5 minutes
### Investigation
1. Check current request rate: `curl localhost:9090/metrics | grep request_count`
2. Review recent deployments: `kubectl rollout history deployment/mission-planner`
3. Check database connection pool: `psql -c "SELECT count(*) FROM pg_stat_activity"`
### Remediation
- If traffic spike: Scale replicas `kubectl scale deployment/mission-planner --replicas=8`
- If DB bottleneck: Clear connection pool and investigate slow queries
- If recent deploy: Consider rollback per deployment runbook
### Escalation
- If unresolved after 15 minutes, page on-call SRE
- If mission-impacting, notify operations center
---
## Alert: DatabaseConnectionExhausted
**Severity:** Critical
...
What AI-Generated Deploy Documentation Often Misses
- Operational context - Why certain alerts matter more during specific missions
- Tribal knowledge - Workarounds and gotchas known only to experienced operators
- User workflow integration - How the system fits into broader operational processes
- Communication procedures - Who to notify and how during incidents
- Classification-appropriate language - User docs for different clearance levels
Human reviewers must validate runbooks reflect actual operational experience.
Governance Checklist
Before accepting AI-assisted Deploy phase documentation:
- [ ] Runbook procedures tested against actual scenarios
- [ ] Incident response contacts current and verified
- [ ] User documentation reviewed for accuracy and clarity
- [ ] Monitoring dashboards match documented metrics
- [ ] Configuration management records complete
Brownfield Additions
For modernization efforts, Deploy phase requires additional focus:
- User transition guides - Document differences between legacy and modern interfaces
- System cutover procedures - Step-by-step activation of modern system with user communication
- Parallel operations documentation - Procedures for operating both systems during transition