The situation
Model risk management was built for models the quantitative team wrote. The bank now runs assistants in operations, in the contact centre, and in three lines of business, and most of them never entered the inventory.
The second line knows this. The first line keeps shipping. When the review arrives, the gap is not the models — it is the evidence, scattered across shared drives and email threads, with nobody able to say what is linked and what is missing.
Cost is a live question too. The bank cannot tell an examiner what these systems cost or what they return without a record that both lines already work in.
What runs where
API gateway
Every metered call, attributed per request, per line of business.
Managed fleet
One deployed file puts every managed seat on the record, including the lines that never told anyone.
Review Readiness
The assurance workspace tracks evidence against the current supervisory edition. It records work status — evidence linked, a question ready for review, an applicability decision made — and never issues a verdict on the bank’s position.
What Oabo shows in the first 30 days
- One inventory of every AI system in the bank, with a named owner and a lifecycle state on each.
- Evidence linked question by question against the current edition, and an honest count of what is still missing.
- Cost per system and per line of business, out of the single platform bill.
- A full audit trail: who recorded what, when, and the source behind each figure.
What to expect
Two units, two blocks. Cash is money your ledger moved. Capacity is time your people got back. Oabo never adds them together.