A prospective mortgage or financial-services client may begin with what sounds like a simple chatbot question and quickly move into territory where the answer could influence an important financial decision. That makes the boundary between administration, information, guidance and regulated advice central to any AI design. For a small firm, the safest opportunity is often to use AI around the advice process — receiving initial enquiries, explaining approved service information and arranging human follow-up — rather than assuming a conversational system can independently become the adviser.
Define the chatbot's regulatory perimeter before its features
The FCA makes clear that the medium used to provide advice does not by itself remove regulatory considerations; interactive software can fall within relevant advice rules. Mortgage advice can also depend on whether communication goes beyond information and is likely to influence a customer's decision.
Firms should therefore map proposed chatbot behaviour against their permissions, applicable FCA rules and their own compliance advice before deployment.
Keep initial enquiry handling distinct from recommendation
An AI tool can explain how the firm's process works, collect contact details and identify the broad reason somebody is seeking help. Those tasks do not require the system to tell the person which mortgage, investment or financial course of action is suitable for them.
Where a conversation begins to depend on individual circumstances or suitability, route it to an appropriately authorised person under the firm's established process.
Use approved information instead of improvised financial answers
Interest rates, product terms, eligibility and regulatory information can change. A general model's remembered knowledge is not a dependable source for a customer-specific financial conversation.
Use maintained, approved material for the information the chatbot is permitted to provide and make stale or unsupported content unavailable rather than letting the model complete gaps.
Be clear about what kind of service the customer is receiving
A customer should not have to infer whether they are reading general information, receiving guidance or being advised. The FCA's current material on automated advice emphasises that firms need to consider the specific obligations applying to their proposition and activities.
Design wording and hand-offs so the system's role is understandable and does not create a misleading impression of personal recommendation.
Collect financial information only when the process needs it
Mortgage and financial enquiries can involve income, commitments, goals and other sensitive personal information. Do not ask for a full financial story merely because conversational collection is technically easy.
Determine what is necessary for initial routing and move detailed fact-finding into the firm's controlled advice process where appropriate.
Give vulnerable or uncertain customers a straightforward human route
Automation should not become an obstacle for somebody who needs clarification, additional support or a discussion that does not fit standard questions. Make human contact visible and preserve useful context when the conversation transfers.
Do not use the chatbot to make unsupported judgements about vulnerability or suitability from conversational cues alone.
Test outputs as regulated customer communications
Before launch, test realistic questions that move from neutral information towards recommendations, including ambiguous requests and attempts to obtain a specific product judgement. Review what the system says, what it refuses and how it escalates.
Ongoing review should involve the firm's compliance and advice owners, not only the technology supplier.
Use AI to strengthen access to advice, not blur accountability
The FCA continues to maintain guidance for automated advice and guidance propositions, including mortgages and other financial-services activities. The practical lesson for a small firm is that adding AI does not make existing responsibilities disappear.
A well-designed chatbot can reduce repetitive initial administration and help prospective clients reach the right adviser with better context. The financial recommendation itself should follow the firm's authorised, governed process, with accountable people and controls appropriate to the service being provided.