Case study · built for our own companies
Running a 41-agent AI workforce for a three-company group, zero failed runs in 13 days
41 agents registered, 9 live, 5,450 runs, 0 failures, 13 days (as of July 2026)
Our group runs three operating companies. The production work that keeps them moving, reconciliation, reporting, publishing, intake triage, catalogue upkeep, is increasingly done by a registered fleet of AI agents rather than by people copying data between tabs.
The numbers, dated because undated numbers are marketing: as of 2026-06-30 the fleet had 41 registered agents, 9 live in production and 5 in beta. In the 13 days measured to 2026-07-04, those agents completed 5,450 runs with zero failures.
What “zero failures” actually means
It does not mean the agents are perfect. It means the operations layer around them is doing its job: every agent is registered, every run is logged, inputs are validated before a run starts, outputs are checked before they land anywhere real, and anything that cannot pass checks stops and asks instead of guessing. Failure is designed to be loud and upstream, not silent and downstream.
That layer, not any single clever agent, is the asset. Agents are cheap to write. An operations discipline that lets you run thousands of unattended executions against real business data is not.
The two-human rule
Today, two humans architect and approve; agents do the production work. Every agent ships with an owner, a scope, and a kill switch. The ambition, stated plainly on our About page, is a fully agentic firm, and we move items from the human column to the agent column publicly as they earn it.
What this proves for clients
The Fractional AI Officer retainer is this operating model, sold. Roadmap, monthly shipped automation, monitoring, weekly written brief. We are not proposing to learn it on your business; we are proposing to install what already runs ours.
Want the same discipline around your operations? Start in writing.