Since 2023
The first university lab dedicated to agentic finance
Engineering the agentic future of finance
City Agentic Finance Lab (CAFiL), part of City St George’s, University of London, helps banks, insurers, payment firms and regulators decide where AI agents add value, test them safely and govern them with confidence.
What we do for organisations
Advisory and Consultancy
Independent advice on where AI agents create value, how to build and govern them, and how to show supervisors that they are under control.
Partnership and funded programmes
Fund a programme, co-fund a pilot or commission a report. CAFiL designs, runs and publishes the work.
Executive Training and Education
Programmes that give leaders in financial services the understanding they need to decide on, govern and oversee AI agents.
Latest Experiments
Reproducible experiments on how AI agents behave in financial settings, from sandbox simulation with no live money to controlled pilots.
Why agentic finance, why now
AI that acts, not just answers
AI agents are beginning to research, decide and pay on behalf of customers and firms. Finance must learn how to use them well — and when not to.
Evidence
In March 2026, Santander and Mastercard reported Europe’s first live end-to-end payment executed by an AI agent — a pilot, not yet a commercial service.
Source: Mastercard Europe newsroom, March 2026
Money is becoming programmable
Instant payments, stablecoins, tokenised deposits and digital currencies are creating rails that software agents can use directly.
Evidence
The FCA published final rules for the UK’s cryptoasset regime in June 2026; they apply from October 2027.
Source: FCA, Cryptoassets, June 2026
Trust has to be engineered
Autonomy without accountability will not last. We design, test and explain agentic systems so that firms, regulators and customers can rely on them.
Evidence
Gartner forecasts that over 40% of agentic AI projects will be cancelled by the end of 2027, because of rising costs, unclear business value or inadequate risk controls (forecast published June 2025).
Source: Gartner press release, June 2025
Invest in CAFiL’s programmes
Funding partners and sponsors support CAFiL’s research and education programmes. Agreements are contracted through City St George’s, University of London; sponsors may comment on drafts but cannot veto findings, and every output discloses its funding.
Latest Experiments
Each experiment starts from a stated question. It runs first in a sandbox with synthetic customers and markets and no live money, then, where useful, as a controlled pilot with a partner. Results are published, including what fails.
- Retail payments and checkout Payments and commerce
- Consumer lending and credit decisions Lending and banking
- Wealth management and financial advice Investment and markets
- Insurance underwriting and claims Insurance and customers
- Regulatory compliance operations Regulation and assurance
- Agent-to-agent markets and market design Investment and markets
Agentic Finance Outlook 2027 — published December 2026
Ten predictions for the year ahead, with the evidence behind them. Free, no registration.
What’s inside
- Ten predictions for 2027, each with a confidence level and the indicator that will score it in December 2027
- Highlights from our Evidence Register, with every deployment graded announced, piloted, live or at scale
- What works and what fails, from the evidence available
- An outlook for each of our four research themes
- Questions every board should ask about AI agents
- Views from industry experts and our Associates
Discuss your challenge with CAFiL
Tell us about the decision in front of your organisation.
Or call us on +44 (0)20 7040 4963