The role of generative and predictive AI in wealth management

What happened?

In a recent webinar, Kit Ruparel, TCC and Recordsure’s Chief Technology Officer, joined Garry Evans, Chief Product and Commercial Officer, to set out a critical distinction: predictive AI is deterministic and produces the same output every time for the same inputs, making it suited to forecasting, modelling and structured data, with tunable confidence thresholds.

Generative AI, by contrast, is probabilistic. Large language models were designed to be creative, producing varied outputs that can be confident and human-like but sometimes incorrect, a phenomenon Kit described as “confidently wrong”.

Predictive AI can reduce human effort dramatically, sometimes by up to 50%, when automating suitability checks and validating structured data. Generative AI is better suited to summarising information into digestible formats, such as portfolio reports or meeting summaries, rather than generating primary insights or critical decisions.

Why does it matter?

The FCA is increasingly expecting organisations to demonstrate strong governance and control, making oversight essential to any AI strategy. Generative AI’s unpredictability means conversations on sensitive topics need guardrails, safety filters and monitoring by information security teams.

Overreliance on generative AI for tasks such as suitability report generation introduces what the webinar called the “AI Verification Tax”, where staff spend as much time checking AI outputs as they would performing the task manually.

Who is affected?

Wealth management firms using, or considering, AI for suitability assessments, advice processes, portfolio reporting and compliance evidence.

Key risks

• Generative AI producing confident but incorrect answers when used without oversight.
• Overreliance on generative AI for primary insight or critical client decisions.
• AI models “drifting” as business processes, customer needs and regulatory requirements change, requiring ongoing retesting.
• Integrating AI without governance frameworks or information security oversight.

Actions to take

1. Match the right AI type to each task: predictive AI for structured, repeatable work; generative AI for summarising and communicating.
2. Build governance frameworks with guardrails, safety filters and information security oversight.
3. Commit to ongoing testing and adaptation as AI models and regulatory expectations evolve.
4. Provide end-user training so staff understand how AI informs decisions.

Wider implications

The successful application of AI in financial services is as much about planning, oversight and understanding a tool’s purpose as it is about the technology itself. Firms should continuously evaluate new AI tools and integrate them into a long-term compliance strategy rather than treating adoption as a one-off project.

Recommendations

TCC is offering a personalised consultation with its Chief Technology Officer to help firms validate their AI strategy and develop a practical, compliant framework.