In partnership with
AI agents are being deployed at enterprise scale across UK financial services right now. The governance infrastructure required to operate them safely does not yet exist in most firms. This issue sets out what has changed, what the FCA is already signalling, and what needs to be built.
On 22 June 2026, Lloyds Banking Group announced it was recruiting for almost 300 agentic AI roles, had over 700 colleagues already building AI use cases, and had placed an AI financial assistant in the hands of over 500,000 Bank of Scotland customers. Real-time payment decisions are now being made by AI fraud detection agents without a human in the loop.
Lloyds is not an outlier. They are the institution saying publicly what is already happening quietly across every major UK regulated firm.
This is not a story about AI hype or future scenarios. It is about a structural accountability gap that is opening right now, between the pace of agentic AI deployment and the governance infrastructure most firms have in place to manage it.
Turn AI into Your Income Engine
Ready to transform artificial intelligence from a buzzword into your personal revenue generator?
HubSpot’s groundbreaking guide "200+ AI-Powered Income Ideas" is your gateway to financial innovation in the digital age.
Inside you'll discover:
A curated collection of 200+ profitable opportunities spanning content creation, e-commerce, gaming, and emerging digital markets—each vetted for real-world potential
Step-by-step implementation guides designed for beginners, making AI accessible regardless of your technical background
Cutting-edge strategies aligned with current market trends, ensuring your ventures stay ahead of the curve
Download your guide today and unlock a future where artificial intelligence powers your success. Your next income stream is waiting.
LLOYDS BANKING GROUP, 22 JUNE 2026
500,000+
Customers already using the AI financial assistant across Bank of Scotland
300
Agentic AI-related roles being recruited in 2026 alone
1,000+
Total AI roles planned across the Group this year
Why the traditional delivery model breaks down
Every large financial institution has a well-established model for deploying technology. A business function identifies a need. A product team forms, works in Agile sprints, defines acceptance criteria, goes through UAT, seeks Change Advisory Board (CAB) approval, and hands over to the business. Risk and compliance review the change at defined gates.
That model was built for deterministic systems. You define what the system does. You test it exhaustively. You know what it will do in production because you specified exactly that.
AI agents are not systems. They are actors. They reason about what to do within constraints you set. That distinction sounds subtle. Its implications for every function in the enterprise are enormous.
You cannot write meaningful acceptance criteria for a system that responds differently to every novel input. You cannot exhaust the test space for a reasoning actor operating in a live financial environment. A CAB process designed to assess a bounded, describable change cannot adequately govern an agent whose full range of future behaviours is not predictable at the point of deployment.
The entire Agile framework, the Model Risk Management (MRM) validation methodology, the change governance model, and the compliance review process were all designed for a world that is ending.
What the FCA is already signalling
The FCA's Emerging Technology Horizon Scan, published June 2026, is the regulator's first external publication of this kind. It is not regulatory guidance. But it would be a serious mistake to read it as neutral commentary. It is the FCA documenting what it can already see in its supervisory work.
FCA EMERGING TECHNOLOGY HORIZON SCAN 2026
"Agentic systems operate at machine speed. If misaligned or manipulated, they can execute thousands of transactions or decisions in seconds. The velocity of harm outpaces traditional firm oversight, making detection and remediation difficult."
The scan identifies four risk areas that carry direct implications for firms deploying AI agents at enterprise scale.
FCA RISK SIGNAL 01
The proxy economy and the SM&CR accountability gap
The FCA describes a shift to a proxy economy where AI agents act on consumers' behalf. It raises the central question: if an AI agent causes harm, who is accountable? The named SMF holder? The vendor? The product team? The FCA does not answer. That silence is the signal. The regulator is watching whether firms build credible accountability frameworks before it has to impose them.
FCA RISK SIGNAL 02
Narrative laundering and synthetic evidence trails
AI can generate complex synthetic evidence trails to obscure malicious intent. The FCA warns that misconduct may become structurally invisible to both firms' controls and supervisory tools as AI-generated documentation grows in volume and plausibility. Compliance functions built to detect anomaly will face threats designed to look like suspicious perfection.
FCA RISK SIGNAL 03
Named individuals carrying liability for machine-speed decisions
SM&CR places personal accountability on named Senior Managers. The FCA's own document describes agentic systems executing thousands of decisions per second. The named SMF holder remains personally liable for oversight of systems they cannot practically monitor in real time, with documentation trails that are inadequate in most firms today.
FCA RISK SIGNAL 04
Emergent multi-agent systemic risk
When agents deployed across different functions interact, behaviours emerge that no individual product team anticipated. The FCA describes agentic systems that may commit market manipulation not through deliberate intent but through the unpredictable interaction of multi-agent systems. Controls designed for individual deterministic systems cannot address this.
FCA EMERGING TECHNOLOGY HORIZON SCAN 2026
We may be moving towards a world where traditional knowledge systems can be exploited, and ground truth may be inaccessible. If this reality comes to pass, we may need systems of governance designed to protect the verifiability of knowledge.
Five infrastructure layers that need to exist now
The traditional delivery sequence places governance weight in the middle: test before deployment, monitor lightly afterwards. The agentic model inverts this. Governance weight shifts earlier, into design and constraint-setting, and later, into continuous monitoring and revalidation. The five infrastructure layers below are what that shift requires in practice.
1 Agent registry and accountability ledger
A live enterprise record of every agent deployed, what it is authorised to do, the named SMF holder accountable for its oversight, and what the pre-deployment evidence looks like. Not a CMDB entry. A structured accountability document producible to the FCA on demand.
2 Pre-deployment evidence infrastructure
Documented proof, created before the agent goes live, that the named accountable individual understood the decision parameters, challenged the constraint assumptions, and exercised genuine oversight. Not assembled retrospectively. Created at deployment as a first-class deliverable.
3 Behavioural interaction mapping
Enterprise architecture that maps not just data flows and API contracts but the behavioural envelope of agents operating in proximity. What happens when the fraud agent interacts with the payments agent interacts with the risk agent? That combined space needs modelling before deployment.
4 Continuous behavioural monitoring
Real-time monitoring that detects behavioural deviation from intended parameters, not just system uptime. An agent running but behaving outside its validated envelope is not a healthy system. Traditional infrastructure monitoring does not capture this.
5 Periodic revalidation framework
Agents that learn, or that operate in changing environments, may drift from their originally validated behaviour. A revalidation cadence linked back to the pre-deployment evidence standard ensures SMF oversight is re-exercised as the agent's context evolves.
The scaling problem compliance teams are not yet talking about
Here is the problem that will become acute faster than most compliance functions are expecting.
As agentic AI deployment accelerates across the enterprise, every deployment requires accountability documentation. Every deployment requires a named SMF holder to have exercised demonstrable oversight. In a large bank deploying multiple AI systems simultaneously across operations, fraud, reconciliation, customer service, risk, and trading functions, the volume of deployments will outpace any compliance team's capacity to review them manually.
The compliance function faces an impossible choice. Scale headcount to match deployment pace, which no firm has budget for. Or find a way to produce audit-ready accountability evidence at deployment speed without adding resource.
Lloyds has 700 colleagues already building AI use cases and 300 agentic roles being filled right now. Every one of those deployments is a potential accountability event under SM&CR. The compliance team reviewing them is not growing at the same rate.
That gap is not a future problem. It is a present one.
Pre-deployment evidential accountability infrastructure for FCA-regulated firms
BuildMoat sits at the deployment gate in the agentic world. We produce the structured, audit-ready evidence packs that prove named SMF holders exercised genuine oversight before an AI agent was authorised to act on behalf of customers.
For compliance teams facing a scaling challenge, BuildMoat is how you maintain accountability standards across multiple simultaneous agent deployments without adding headcount. For the named SMF holder, BuildMoat is the documented answer to the question the FCA will eventually ask.
We work with compliance teams, risk functions, and named SMF holders at FCA-regulated firms to operationalise pre-deployment evidence infrastructure as a built-in part of the deployment process, not a retrospective exercise after something goes wrong.
The Lloyds announcement is the start. Every major UK regulated firm is either here already or heading here fast. The compliance teams and named SMF holders that build the evidence infrastructure now will not be scrambling when the FCA comes looking.
If this issue raised a question your firm is grappling with, reply directly to this email. I read every response.
Emeka Ogbonnaya
Founder and CEO, BuildMoat
