Skip to content

Workflows

A workflow (WF-0xx) chains specialized agents into an end-to-end deliverable. Each step is gated: the run only advances when its blocking checks pass.

Proven workflows

Nine workflows have been run live, end to end. WF-001 ran against catalog v4.0.0, WF-002 and WF-003 against v3.27.0, and WF-004, WF-005, WF-006, WF-007, WF-009 and WF-010 against v4.1.0:

Workflow Use case Steps Status Result
WF-001 AI product scoping 3 completed 3 / 3 passed
WF-002 SAFe agile delivery 5 completed 5 / 5 passed
WF-003 RAG customer-support chatbot (insurer) 7 completed 7 / 7 passed
WF-004 AI consulting engagement 6 completed 6 / 6 passed
WF-005 Strategic intelligence & growth 3 completed 3 / 3 passed
WF-006 Pre-sales / commercial proposal 6 completed 6 / 6 passed
WF-007 Mission onboarding — Day 1–5 4 completed 4 / 4 passed
WF-009 IT / AI recruitment 6 completed 6 / 6 passed
WF-010 Project post-mortem 4 completed 4 / 4 passed

The tenth workflow, WF-008 (AI Act / GDPR compliance audit), was run live against v4.1.0 and returned for rework by its counter-review gate, so it is not an end-to-end proof — with it, all 10 catalog workflows have traversed live (nine completed, one returned). WF-009 was proven completed on a synthetic, provided candidate pool (fictional by design), not autonomous sourcing; the same shortlist gate halted a no-candidate variant, so the two WF-009 traces together show the gate discriminates. Full per-step verdicts for every run are captured as JSON — see Live proofs.

Detailed example — WF-003

Use case: a RAG customer-support chatbot for an insurer, with answers sourced from a policies/claims knowledge base.

The workflow runs 7 steps, each delegated to a dedicated agent from catalog v3.27.0:

  1. AGENT-FINANCIAL-ANALYST — business case & TCO
  2. AGENT-PROMPT-ENGINEER — prompt design
  3. AGENT-AI-ARCHITECT — solution architecture
  4. AGENT-DEV-PYTHON-IA — implementation
  5. AGENT-QA-AGILE — quality assurance
  6. AGENT-DEVOPS-CLOUD — deployment
  7. AGENT-SECURITE-IA — security review

See the verified end-to-end run in Live proofs.