FCA Links Agentic Finance To Tokenisation And Third-Party Risk
FCA chief executive Nikhil Rathi said AI is moving financial markets faster than traditional rulemaking, with agentic systems, tokenisation and third-party model dependence reshaping supervision.
FCA Puts AI Supervision On A Faster Clock
The UK Financial Conduct Authority is recasting AI in finance as a supervision, competition and resilience problem, not only a technology adoption story.
Chief executive Nikhil Rathi used a 24 June 2026 speech at techUK's Agents of Change event to argue that financial services will sit at the centre of the UK's AI economy because the sector supplies capital, infrastructure and trust.
Rathi put AI adoption above 80% across financial services firms.
He framed the next stage as scale, with technology moving faster than the assumptions behind many market and regulatory frameworks.
The FCA's emphasis is important for banks, payment firms, asset managers and market infrastructure providers because the regulator is tying AI use to accountability, competition and operational resilience.
Rathi said accountability for regulated activities and outcomes must remain clear when firms use systems that support decisions, coordinate activity or transact.
Agentic Systems Meet Tokenised Markets
Rathi identified agentic systems and tokenisation as two large scaling opportunities.
In retail markets, he said agentic systems could support bill management, personalised investment strategies and lower friction.
In wholesale markets, he pointed to liquidity management, trading workflows and other market functions.
The regulator did not present agentic finance as ready for unchecked delegation.
Rathi said investors will be wary of handing important decisions to systems they do not understand, and he tied consumer confidence to human oversight and clear accountability.
Tokenisation was the second pillar.
Rathi said tokenisation could lower costs, reduce risk and unlock services by creating more automated and programmable infrastructure for agentic finance.
He said banks are already piloting tokenised deposits, including use cases that could reduce friction and fraud risks in home buying.
The speech also pointed to a concrete market step: Baillie Gifford and Bank of New York Mellon received FCA approval for the UK's first natively tokenised authorised fund.
Rathi said the FCA and the Bank of England have set out a direction for tokenised wholesale markets through a joint call for input that closes next week.
Cloud, Models And Fraud Become Resilience Issues
The resilience section moved the speech beyond productivity claims.
Rathi said financial services firms increasingly rely on cloud providers, model providers, data providers and other parts of the AI stack.
He linked those dependencies to systemic resilience, market integrity and financial crime.
The speech cited UK Finance's Annual Fraud Report, saying the UK lost almost £1.3 billion through payment fraud last year and that two-thirds of authorised fraud cases came through social media sites and messaging platforms.
Rathi also said 98% of operational incidents reported to the FCA last year related to technology and cyber issues.
Those figures put AI governance inside a broader third-party and platform-risk agenda.
Rathi said boards and leadership teams must understand the risks, map and govern model-provider and third-party dependencies, and treat the Critical Third Parties regime as more important.
FCA Tools Move Toward Real-World Testing
The FCA is also testing its own operating model.
Rathi said the regulator is exploring agentic AI as a first responder for wholesale-market monitoring, using large data sets of a billion rows per day alongside supervisory judgement to tackle market abuse faster.
He pointed to the Supercharged Sandbox, where firms can test solutions with real-world data, compute and systems with partners including Nvidia and soon Google.
He also cited the AI Lab for payments and e-commerce, a dedicated Agentic Academy and an AI Consortium with the Bank of England.
For regulated firms, the speech leaves a specific compliance burden rather than a broad AI slogan: AI projects will need clear accountability, mapped model and cloud dependencies, fraud controls, competition-law awareness and evidence that human oversight still governs regulated outcomes.
