AI agents in finance operations: what the 2026 research says, and what to do about it
Six findings from Databricks and BCG, read from the finance owner’s seat. Why pilots stall, why governance is what gets agents into production, and the four things to insist on before any agent touches your ledger.
1
The six numbers. Adoption, pilot failure, governance, impact definition, finance in the room, and the gap between pilots and real change, each with its source and date.
2
Why a smaller firm can out-execute a larger one, and the five priorities BCG sets for doing it, read for a UK services firm.
3
The method. A signed mandate, a confidence score, a tamper-evident record and a named owner, and how sixty days works as investment discipline.
95%of generative AI pilots never reach production. MIT NANDA (2025), cited in Databricks (2026)
12xmore AI projects reach production with active governance. Databricks (2026)
40%more likely: high performers put finance inside AI governance. BCG (Sept 2026)
Anton Russell, Founder and CEO, MandateEx-RAF Intelligence, ex-EY. Mandate deploys accountable AI agents for finance operations, starting with accounts payable.
