News of the Week

Agency-led wealth briefing — 26 October 2026

The material development is the FCA’s move from broad AI principles to observable operating practice. Its message is clear: firms remain responsible for outcomes produced through AI, including autonomous agents.

1. The FCA has defined practical expectations for financial AI

Status: Confirmed supervisory guidance; it creates no new rules. Published 19 October, updated 20 October.

The FCA’s new good and poor practice review covers generative and agentic AI under existing requirements including Consumer Duty, governance, operational resilience and accountability.

Good practice now includes:

• defining the consumer need before building the tool;

• deciding where AI may support, automate, abstain or escalate;

• testing outcomes across vulnerability and financial-capability scenarios;

• explaining what the AI can and cannot do;

• retaining prompts, outputs, model versions, filters and escalation records;

• giving human reviewers the knowledge and authority to intervene;

• monitoring outcomes continuously rather than approving a system once. FCA

The FCA also warns that “human in the loop” can become a box-ticking control. Effective oversight requires actual capability, authority and escalation routes. Smaller organisations may use proportionate governance, but still need named owners, documented decisions and fallback arrangements. FCA

Why this matters to AoLP

This closely supports the Academy OS architecture: bounded tools, approved knowledge, visible limitations, escalation to a qualified person and an enduring decision record. It also validates financial capability as something AI testing should measure.

The distinction AoLP can own is:

Institutional governance asks whether the firm can control the system. Agency-led governance also asks whether the person can understand and control what the system does for them.

Action: create an Academy OS Agent Register. For every app or agent, record:

• intended user outcome;

• information and systems it can access;

• whether it supports, automates, abstains or escalates;

• prohibited actions;

• test cases covering capability and vulnerability;

• accountable owner and human escalation route;

• model and knowledge-source versions;

• failure, fallback and withdrawal arrangements.

2. The FCA has left the personal financial-capability gap open

Status: Confirmed FCA policy direction, expressed in a speech rather than a rule.

On 19 October, FCA Chief Data, Information and Intelligence Officer Jessica Rusu said four in five less-experienced investors aged 18–40 had used AI for investing, while only 40% recognised that general-purpose AI advice carried no regulatory recourse.

The FCA is calling for clearer point-of-use disclaimers. More significantly, it will not implement the Mills Review proposal for a new public-interest financial-capability tool. It will instead extend InvestSmart, Firm Checker and work that helps investors detect scams, verify information and challenge unreliable AI output. FCA

Why this matters to AoLP

This leaves space for Academy OS to become personal capability infrastructure:

• helping people question and verify AI output;

• maintaining their own evidence and decision history;

• distinguishing education, guidance, coaching and regulated advice;

• knowing when professional help is required;

• preparing the person to use episodic expertise effectively.

That is a stronger proposition than simply offering another financial chatbot.

Action: make these capabilities explicit in Ask My Adviser and the Total Wealth Planner journey. At the point of use, show the source, limitations, regulatory status, next verification step and route to human support.

3. Agent governance is moving towards runtime controls

Status: International cooperation and emerging practice; no new UK rule.

On 20 October, HM Treasury and the Monetary Authority of Singapore discussed Singapore’s Safeguards for Agentic Finance at Runtime, alongside the FCA’s AI Lab and joint work on AI taxonomies. The SAFR approach concerns how agents carry out financial tasks safely and reliably while they are operating. GOV.UK

The FCA separately said it intends to work on a framework for agent identity, delegated authority, accountability and execution, with continuous, conditional assurance for agentic systems. This is a direction of travel rather than completed policy. FCA

Implication for Academy OS: permissions should be visible and revocable. An agent should know whose interests it serves, what it may do, when consent expires and when it must stop.

No consequential update this week

• Professional standards: no material new financial-planning qualification or conduct standard located.

• Business models: no confirmed market change strong enough to alter the Academy OS or Total Wealth Planner thesis.

• General AI capabilities: no provider release this week that materially changes the regulated financial-planning use case.

• New UK AI legislation: none. The FCA continues to apply existing outcome-based rules to AI.

Priority

Build the Academy OS Agent Register and turn it into a public-facing Agent Charter. This would translate the FCA’s supervisory language into a client promise: the person can see what the AI knows, what it may do, where its authority ends and how to obtain human help.