Source Insurance, a UK home and landlord insurance intermediary with over 30 years' experience, is developing a conversational AI interface for home and landlord insurance referrals. The platform has been co-developed openly with advisory firm Connect Mortgages. Iterative testing, deliberate challenge and structured scepticism from advisers have shaped each development cycle. The goal is to earn customer trust one exchange at a time, replacing rigid online portals with a guided, natural conversation.
Source Insurance, a UK home and landlord insurance intermediary with over 30 years' experience, identified a specific failure point in general insurance referrals. The issue is not the handover itself but what follows it. Traditional portals and rigid online forms cause customers to disengage at precisely the moment they most need guidance. Chris Lynch, IT director at Source Insurance, put it directly: 'Referral solutions never quite solved the hardest part of a GI referral. It's not making the referral. It's keeping customers engaged once the adviser has handed over. You can't meet the customer's need if the customer has logged off.' The conversational interface targets that drop-off by replacing portals with a guided back-and-forth exchange that keeps customers present throughout the quoting process.
Instead of asking customers to translate their circumstances into rigid pre-formatted fields, Source Insurance's interface guides them through a home insurance quote via a back-and-forth exchange. It moves one question at a time in chronological order, rephrases when something is unclear and adapts based on each response. Messaging has become second nature for many consumers, and interfaces such as ChatGPT, Claude and Gemini have normalised AI conversation as a format. Source's intention was to make the insurance quoting process feel equally natural. Lynch described the relationship underpinning the build: 'Connect Mortgages are an integral partner for us as we continue to beta test and refine.' That partnership has been the mechanism for testing whether the experience genuinely feels conversational rather than merely formatted as one.
Source Insurance took the position that assumptions about where AI should and should not act must be tested by the people who use the system every day. Building without external input would have produced a platform shaped entirely by internal assumptions. Liz Syms, CEO of Connect Mortgages Limited, described what the collaboration has looked like in practice: 'We've been through multiple test cycles with Source. We've delivered our feedback, talked through solutions and they've gone away, worked on improvements and delivered a new version for us to test again.' That repeating cycle of challenge, rework and retest has replaced a standard closed-door development model. Each iteration is driven by real adviser experience rather than by what the development team predicted advisers would encounter.
Connect Mortgages has acted as a deliberate stress-tester throughout the project. The partnership has focused specifically on finding failure points rather than confirming what already performs well. Lynch explained what that scrutiny produced: 'Together, we've deliberately pushed the system to its limits. We've tried to find the moments where it falls short, where the experience feels unnatural and where human involvement is needed. Those challenges have helped us build stronger boundaries and helped us understand where to add hard-coded guardrails to continually improve the experience.' Those guardrails now determine when the AI responds, when it asks for clarification and when it stops and passes the customer to a human adviser. The boundaries were drawn from observed failure, not from theoretical design.
Source Insurance treats the boundaries of AI involvement as trust decisions rather than purely technical ones. The development team worked through four core questions: where should the AI respond; where should it ask for clarification; when should it acknowledge uncertainty; and when should it stop and bring in a human adviser. Syms reinforced why adviser input into those decisions matters for practical adoption: 'Advisers having direct input into a solution they will be using is essential for adoption. They are the ultimate test group.' Each test cycle has sharpened where the system draws those lines. The result is a platform whose limits are understood before customers reach them, rather than discovered midway through a quote.
Enthusiastic early adopters rarely surface the moments where a system breaks down. People who hold a healthy wariness do. Source Insurance has treated scepticism as a practical design tool throughout the project. Lynch noted that every round of testing challenges another assumption and exposes where the experience feels forced rather than natural. Sceptics ask harder questions, hold higher standards and push development teams to earn confidence through performance rather than assume it through claims. Source's own starting point was deliberate wariness, and the platform has been shaped by that discipline from the outset. The hard-coded guardrails now built into the system exist specifically because external testers refused to accept anything less than consistent, reliable behaviour at every point in the exchange.
Source Insurance is drawing a clear lesson from its development process: adoption will not come from persuading people that AI is reliable. It will come from building systems that respect the reasons people are wary and earn confidence through repeated, consistent performance. Lynch's framing captures the logic: 'You can't meet the customer's need if the customer has logged off.' The same principle applies to trust. If customers and advisers disengage because an experience feels unreliable, no underlying technical capability can recover the relationship. Source holds a 5-star Defaqto accredited panel rating and a 5-star Trustpilot rating. It is applying those same standards of scrutiny to its AI platform, treating adviser confidence and customer retention as the real measures of whether the technology has succeeded.
Chris Lynch: IT director, Source Insurance; led development of Source Insurance's conversational AI interface for home and landlord insurance referrals; responsible for the technical design and guardrail architecture of the platform.
Liz Syms: CEO, Connect Mortgages Limited; co-led beta testing of the Source Insurance conversational AI interface; principal external challenger throughout multiple iterative development cycles.