The UK's Financial Conduct Authority (FCA) has released a study predicting that autonomous, agentic AI will have a "seismic" impact on financial services by 2030. The Mills Review, commissioned by the FCA, found that 81% of firms are adopting AI at some level, with 40% implementing scaled or transformational stages. The shift from human-led episodic work to AI-enabled autonomous agents is already underway.
As autonomy grows, accountability becomes harder to trace. The study warns that "a useful answer is not enough if the basis for it cannot be reconstructed." For investment firms, the need for grounded, traceable, auditable AI has never been greater.
Key Findings and Recommendations
The review details fundamental risks of mainstream AI adoption: hallucinations, model inconsistency, response drift, and failure to produce regulated, traceable responses. It outlines several recommendations for the FCA Board:
- Secure the Regulatory Perimeter: Launch an immediate review into the scale and impact of general-purpose Large Language Models (LLMs) that operate outside current regulation.
- Strengthen Coordination and Oversight: With no dedicated AI regulator, tighten coordination with national and international partners across model, cloud infrastructure, and distribution providers.
- Monitor Autonomous Models: Provide clarity on accountability, governance, and consumer protection frameworks to support dynamic model governance and end-to-end controls across the AI lifecycle.
- Scale the FCA AI Lab: Build a structured capability to assess AI models and systems before they are embedded in core operations, supporting "responsible growth" and addressing explainability challenges.
- Foundations for Agentic Finance: Establish trust frameworks, permissions, and authentication for autonomous AI agents to operate safely.
- Agentic Supervisory Model: Deploy AI-enabled tools for authorization, supervision, and real-time monitoring to detect system-wide risks invisible at the individual firm level.
- Public-Interest AI Service: Convene the development of a free, inclusively designed AI-enabled service to give consumers reliable financial information and guidance.
The Challenge Firms Face
Enterprise AI integration is inevitable, but general model use is prone to hallucinations, insecure data handling, and illogical conclusions. The Mills Review states: "General-purpose and frontier AI models can introduce challenges including opacity, model drift, data bias, hallucinations and emergent behaviours."
According to broker research from AlphaSense, 87% of UK business leaders anticipate spending 10% or more of their total budget on AI over the next 12 months. The question for investment firms is no longer whether to adopt AI, but how to support enterprise integration that keeps pace with evolving technology and meets compliance requirements for highly regulated work.
The Gold Standard
Vertically integrated intelligence systems unite trusted content, retrieval capabilities, context, orchestration, and synthesis in one place. This approach ensures outputs are verified, defendable, and citable, protecting against "black box" outputs identified in the FCA study.
Each component matters. Models need access to relevant information, tools, and institutional knowledge. A context graph, built on a knowledge graph that resolves entities and relationships across millions of documents daily, structures data before any query arrives. For example, an analyst querying Apple's supply chain exposure can distinguish between Apple the company and its suppliers. Decision quality is shaped by the right evidence, not just any available evidence.
As the demand for trustworthy AI grows, many firms are turning to specialized platforms that deliver AI for Finance with built-in auditability. This gold standard for enterprise adoption removes risk by bringing all components together. The result is a faster, simpler, and less costly path to a better answer, with token efficiency built in.
A Trusted Path Forward
With AI becoming more autonomous, firms are at a critical juncture. The FCA study makes clear that grounded, traceable, auditable AI is essential. Platforms that unite end-to-end capabilities-content, retrieval, context, orchestration, synthesis, and auditable outputs-offer the verifiability financial services firms expect while keeping them competitive.
Why this matters for finance professionals
The FCA's recommendations signal that regulators will increasingly demand transparency and accountability in AI systems. For finance professionals, deploying AI without clear audit trails and explainable outputs is a direct compliance risk. The study's emphasis on dynamic model governance and end-to-end controls points to a future where AI tools must be both powerful and provably trustworthy. Choosing AI infrastructure that can reconstruct every answer will become a baseline requirement, not a differentiator.
Your membership also unlocks:
