In February 2026, MoneySuperMarket became the first UK price-comparison brand to ship inside ChatGPT. Aviva followed in March. MoneySuperMarket's chief executive told the press that price-comparison sites weren't being replaced — they were "moving into the plumbing." He's right. And most of consumer finance hasn't caught up.
This is what the next decade looks like.
The chat layer is the new front door.
The funnel for finding a financial product no longer starts at Google. It starts in a chat window.
EY now reports half of UK consumers use AI for financial decisions. Apple has put a 3-billion-parameter language model on every recent iPhone — roughly 2.2 billion devices walking around with a personal LLM in their pocket. The newly announced Apple Intelligence Pro tier goes further: it sells agentic actions — Siri completing multi-step jobs across apps, like "find the flight in my email and book the Uber for that time." Gartner has predicted that 40% of all financial-services firms will be using AI agents by the end of 2026.
The same pattern is coming to financial products. The customer journey starts as a sentence, not a search: "I need a £15k loan for a kitchen — what's the cheapest five-year option?"
That sentence used to land on Google, then on a comparison page, then on a quote form. The new path is shorter. It lands in a chat — and looks something like this:
ClearScore's CEO, Justin Basini, put it cleanly when launching the world's first protocol for agentic credit broking earlier this month: "users will, of course, begin their financial journeys through AI assistants and applications, much like they did many years ago through the internet." The internet metaphor matters. Anyone arguing this won't happen is a 1998-era retailer telling you e-commerce is a fad.
Search hands you three words. The AI hands you a customer.
This is the bit most boards are missing. Google's footprint into a lender's funnel was always thin: a search string, a referrer URL, maybe a cookie. Everything that mattered — intent, context, affordability, identity — had to be re-collected through forms, declarations and KYC. Every step was friction, and most of the drop-off in finance journeys happens precisely because the journey starts that cold.
A personal AI's footprint is structurally richer. By the time a customer says "I need a £15k kitchen loan", the assistant already knows the context behind the question — the conversation that led to it, the budget the customer mentioned last week, the bank account they linked, the calendar event that triggered the kitchen project, the affordability picture they've already shared. When the AI hands the customer to a regulated broker, it's not handing over a search string. It's handing over a structured payload — pre-validated, pre-contextualised, pre-consented.
That's the seamless handoff. The opportunity isn't that consumers will start asking finance questions in chat — they already do. The opportunity is that the broker on the receiving end of that chat now has more, better, faster context than any Google-led journey has ever delivered. The brokers and lenders that lead with AI — that build their stack to ingest, interpret and act on that rich payload from the moment the customer speaks — will win the next phase of price comparison as it evolves into AI.
What this kills.
The classic aggregator model was built on one core assumption: that the customer journey starts on a results page that the aggregator owns. PPC at the top, sponsored placement in the middle, lead capture down the side. The whole P&L flows from being the first page the customer sees.
When discovery moves to chat, that page becomes optional.
Three futures play out for today's aggregators. A few will become the broker-inside-the-AI — MoneySuperMarket and Aviva are already trying. A few more will become a backend rail: useful, invisible, low-margin. Most will be squeezed in the middle, fighting for the same dwindling traffic that ChatGPT, Claude and Siri are quietly disintermediating.
What lenders and brokers need to do about it.
The SEO playbook is over. The new playbook is being legible to an AI.
Call it visual AI mechanics — the discipline of showing up well when an LLM presents your product to a customer. Structured product data. Clean, well-documented APIs. MCP-style integrations that let an AI agent query your rates and apply for a quote inside a chat. Rich product cards that render properly when ChatGPT or Apple Intelligence surfaces them. A live data layer that knows what you know — not what the comparison-table cache knew yesterday.
For the boards in lending, broking and insurance, that means a real shift in where strategy spend goes. The next pound is not best spent on more PPC. It's best spent on the operating layer that makes your product visible — and accurately representable — to an AI.
The protocols are already arriving.
Here's the part most lenders haven't clocked: the rails are being laid right now, and they're being laid in the open.
In April 2026, ClearScore Group open-sourced the Agentic Credit Broking Protocol (ACBP) — what it called "the world's first standard for agentic credit broking." The framing in their launch document is the clearest articulation yet of the structural problem facing the industry: "AI assistants break that model. A user may begin their credit journey through a chat assistant, a financial app, a comparison tool, or an embedded assistant in another service. The regulated firm no longer controls the first interaction surface."
ACBP's answer is to let "interaction and responsibility travel separately." An AI agent can have a natural-language conversation with a customer about their income, outgoings and existing debt; hand a structured payload to a regulated broker; let the broker source offers from lenders; and route the customer back through the AI to complete the unregulated parts of the journey. The agent never has to become a regulated entity. The broker keeps the licence and gains an audit trail.
Alongside this, more general agent-orchestration frameworks are emerging. Open-source projects like ClearCoreAI are building the auditing and governance primitives — manifests, policies, traceability of every agent step — that any regulated use case will eventually need. The protocols and the substrate are arriving in parallel.
The platform underneath.
Protocols are necessary but not sufficient. ACBP says how an AI agent talks to a broker. It doesn't tell a lender how to be the broker.
That's where Credit Canary fits.
Credit Canary is the operating platform purpose-built for this future. Six modules — Originate, Unify, Decide, Pay, Act, Performance — wired together with their corresponding agentic counterparts. Origination data normalised across channels. A unified customer view that holds up under regulator scrutiny. Decisioning with full evidence trails — exactly what ACBP-style protocols expect to receive when an AI agent hands off a customer journey. Payment and engagement flows that work whether the customer arrived from a website, a comparison page, or an agent inside their personal AI.
It's the kit a lender or broker actually needs to ship into the chat-first era. Origination through performance, all in one stack, all agent-ready, all auditable from end to end. The protocols arriving in 2026 are inert without a platform that can actually execute on them — and Credit Canary is the platform that lights them up.
If you're a lender thinking "the protocols are real, the regulator is paying attention, but our origination stack runs on five vendors and three spreadsheets" — that's exactly the gap Credit Canary closes.
The regulator is already moving.
The FCA isn't asleep on this either.
In January, the regulator launched the Mills Review — a long-term study of how AI will reshape retail financial services, with findings due to the FCA Board in summer 2026. Sheldon Mills, the FCA's executive director leading the review, warned that "advanced, multimodal and agentic AI systems could reshape market dynamics, alter how financial products are designed and distributed, and transform how consumers engage with firms" — and that the technology "could also potentially reduce consumer agency and introduce new forms of market concentration or systemic vulnerability."
In March, the FCA's perimeter report flagged a specific edge of its remit: the "rapid growth of general-purpose AI tools offering financial advice or recommendations." The FCA's own guidance is blunt: outputs from ChatGPT and Claude aren't covered by the Financial Ombudsman Service or the FSCS.
Then on 25 March 2026, the FCA published its Payments Regulatory Priorities report, with what's so far the regulator's most concrete commitment to the technology: it intends to rewrite payments rules to accommodate agentic AI. Eight more firms are now in the second cohort of AI Live Testing.
What does this mean? The chat layer is happening regardless. The regulated transaction layer underneath it is still up for grabs. And the FCA is paying attention.
Who holds the licence — and why it matters.
This brings us to the more interesting strategic question. The chat layer is happening. The protocols are arriving. The regulator is watching. So who actually holds the licence on the other side of the chat window?
Anthropic has moved aggressively into financial services. Claude for Financial Services launched in 2025, with data partnerships that read like a Bloomberg roll-up: S&P Global, LSEG, Third Bridge. AIG used Claude to compress its underwriting review timeline by 5x. Norway's sovereign wealth fund reported 213,000 hours saved. In April 2026, Piraeus Bank and Accenture launched an Anthropic-powered AI Hub for the Greek banking sector. The foundation-model layer is already deep inside enterprise finance.
Could a foundation-model company go the next step and take the FCA permission itself? Theoretically. Credit broker. Insurance intermediary. Payment services firm. The unit economics of a regulated transaction are better than per-token pricing, and an AI provider sitting on hundreds of millions of consumer conversations is well placed to monetise them.
In practice, that's a path very few AI providers will actually want to walk. The compliance overhead, the consumer-duty exposure, the FOS and FSCS perimeter questions, the conduct-risk implications — none of it fits a foundation-model business model. What every AI provider actually needs is a regulated counterparty: a partner that holds the permission, owns the consumer relationship in regulatory terms, and makes sure the financial journey can be safely served inside chat.
That's the role Credit Canary is built to play. The platform sits exactly where the licence has to sit — at the regulated, audited handoff between the AI conversation and the executed transaction. ACBP-style protocols assume a regulated broker on the receiving end of every chat journey. Credit Canary is that broker. Originate → Performance is wired end to end so the AI can hand off into a stack that actually carries the consumer-duty load — capturing consent properly, evidencing affordability, treating vulnerable customers fairly, and producing a complete audit trail when the regulator asks how the journey unfolded.
The question for boards is no longer "will Anthropic take the licence?" It's "who is the regulated platform that lights up the chat journey for our customers — safely, auditably, and at speed?" If your origination, decisioning and engagement are already on an audit-grade, agent-ready platform, the AI agents will route through you. If they're not, they'll route around you.
What this means for your board.
If your product can't be cleanly described, priced and presented to an AI agent, it won't be sold by one.
The boards still defending SEO budget for compare-and-click pages are fighting the last war. The ones investing in the operating platform underneath their products — origination data, a unified customer view, decisioning that holds up under scrutiny, agent-readable APIs, engagement flows built for the chat era — are fighting the next one.
That's the work. It's not glamorous. But it's the difference between being a result the AI surfaces — and being invisible to the next generation of customers.
The aggregator is moving into the plumbing. The broker is moving into the chat. And the lenders that win the next decade will be the ones who, today, decided to build their stack on rails that an AI agent can actually use.
Credit Canary is built for that fight.
← Back to News & Opinion