Operating Model Zero
I know operating models from the inside — global markets front to back. When AI arrived, I didn't reach for a framework — I built a working multi-agent system and ran the middle-office reconciliation loop through it, break to resolution, on representative data. Operating Model Zero is the frame; the system is the proof: agents don't bolt onto the model, they rebuild it.
Take today's middle office. Add a copilot to one step. Book the saving. The org chart, the controls, the cost base — unchanged. The existing model is slightly faster and slightly more efficient. The shape is the same.
Rebuild from zero. Redraw the function with agents native to the design: how many people, in what roles, with what controls. Zero-based is the design target; you reach it by sequencing the transition under a live book and live controls — not by stopping the business to start over.
The same work, a different shape.
The work is redistributed. Agents absorb the high-volume, deterministic load; the human layer gets thinner and more senior — judgment, exceptions, and the decisions that carry liability.
But accountability doesn't thin out with the headcount — it concentrates. When an agent books the match and chases the amendment, the task has moved, but the liability hasn't: it still sits with a named person, answerable to the board and the regulator. So the question underneath the whole redesign is a governance one — who is the accountable executive when agents do the work? — and as the human layer gets thinner, the answer lands on fewer people, higher up. Deciding who owns which decision, and where the buck stops when an agent is in the loop, is the core of the operating model, not a footnote to it. Zero-based means clean-sheet, not unmanned.
A worked example, end to end.
Operating Model Zero, built and running — a working multi-agent system for FX reconciliation. It runs on representative trade data with a simulated counterparty, demonstrating the full break-to-resolution loop end to end: it parses a counterparty confirmation into structured economics, matches it against the book, detects the break, then works it to resolution — querying the front office, pursuing a counterparty amendment, and validating the fix. The humans are left to the exceptions and the calls that carry accountability.
Built on a deterministic backbone: the agents make the judgment calls, but the arithmetic and the audit trail are never left to a model. Every action is recorded — and a control you can't audit isn't a control.
Sebastien Bietho. Fifteen years designing and building the operating models behind global markets businesses — target operating models, front-to-back trade flows, and the end-to-end build of new trading platforms and entities — grounded in a decade on the trading floor running securities-finance books with full P&L ownership. He now advises on a board-endorsed regulatory remediation at a major financial institution and leads its design and implementation. Operating Model Zero — and Exhibit 01, the working system behind it — is what happens when someone who builds these models decides to find out what AI does to them.