Those gains may sound operational, but they can improve the quality of the advice process. Where capacity is tight, better organisation can help keep recommendations clear, cases moving and clients informed. For many firms, the best place to start is with low-risk tasks where outputs can be checked easily and accountability stays clear. That might mean using AI to capture action points, compare documents for inconsistencies, or help staff navigate internal procedures.

The FCA’s focus on customer outcomes runs through all of this. Advisers using AI still need to show that products and services meet the needs of the target market, that communications support customer understanding and that support remains appropriate.

A sensible rule is to separate support tasks from advice decisions. If a tool helps organise information, draft routine wording or identify missing data, the risk is usually easier to manage. If it influences recommendations, prioritisation or how options are presented, the bar should be much higher. Advisers need to ask basic questions before using any tool: what is it doing, what information is it using, how reliable is the output, and how easy is it to challenge or correct? If those answers are unclear, that should slow the process down.

Data security also needs to be part of the decision from the start. Advisers handle highly sensitive personal and financial information, taken from fact finds, payslips and bank statements, credit reports, protection needs and any health details shared when arranging life or critical illness cover. Health information is treated as a special category under UK GDPR, which means stricter rules apply to how it is stored and used. This is a particular risk with free or public AI tools, which often use whatever is typed into them to train the tool and are not designed for confidential client data.

Before any information is entered into an AI tool, firms need to be clear on whether that data is being stored, used to train the tool, shared with other companies or sent outside the UK, and these points should be confirmed in writing by the provider rather than assumed. If there is any doubt, information that could identify a client, or that is commercially sensitive, should not be entered into a tool that has not been checked and approved for that purpose. Firms should also keep a record of which tools have been approved and why, so the decision can be evidenced if the FCA asks. AI may save time, but it should not weaken the standards firms already apply to confidentiality, data protection and keeping client information secure.

Accountability stays with the adviser. Responsibility doesn’t move to the technology. Existing rules still apply, so firms need to think carefully about explainability, governance and human oversight. They should be able to explain how AI influences decisions and make sure a qualified person retains the final judgement on advice or recommendations.

This matters because AI can sound confident even when it’s wrong. A polished summary may miss an important qualification, or a drafted paragraph may introduce assumptions that were never discussed with the client. That’s why review can’t be treated as a formality. Someone with the right experience needs to sense-check the output, understand where it could mislead, and be willing to override it.

That has practical consequences in adviser businesses. Drafting tools shape language, summaries can shape judgement, and automated prompts can encourage shortcuts if no one’s checking them properly. Firms need clear oversight over who approved the tool, who reviews output, and whether it’s suitable for client-facing use.

The FCA is creating space for firms to explore AI in controlled settings. Through its AI Lab and Supercharged Sandbox, it has opened routes for firms to test ideas with regulatory support, synthetic datasets and specialist tooling. That gives firms a practical way to experiment without losing sight of governance, accountability and customer outcomes.

That is exactly why firms should be getting involved early. Firms that engage with AI now are better placed to shape tools around real advice needs, build internal understanding and test risks before expectations tighten further. Waiting for perfect clarity is unlikely to help. The better approach is to start small, set boundaries and check outputs consistently.

For advisers, the next step is to be selective and practical. Start with a business problem, not a tool. Use AI where it reduces friction, supports better client outcomes and improves consistency. Keep human accountability clear. Test outputs properly. Make sure staff understand what the tool can and can’t do. Keep client-facing use under tighter control than internal support work. Most of all, treat AI as something that still needs judgement, not something that replaces it.