Always-on agents, each with a narrow brief over one slice of governed data. Every run reasons over real figures — the reconciled FY25 book, or the live governance repository — and writes to the same decision log the Today queue and Activity feed read from, so an agent action leaves an audit trail rather than a claim.
Run one, or run the workforce. Agents are deliberately narrow: the Compliance, Knowledge and Data Engineer agents read the live Postgres governance repository, the rest read the reconciled book. None of them is a general chatbot — that is what Ask AI is for.
The agents above are the orchestration layer. These are the underlying assets — the ones marked Live are real Generali Central systems; the rest are peer-benchmarked targets.