Artificial intelligence (AI) has had quite a year in the mid-market.
It has gone from a curiosity to something that every brokerage firm promotes.
Discussions are currently mostly about efficiency, speed and time saved: tools that build relationships with clients, shorten calls or generate lead follow-ups in record time.
I understand the appeal. Brokers are under pressure, margins are tight and anything that brings back time will be welcome.
AI in lending must be purpose-built and work with our infrastructure
But much of what is currently badged ‘AI’ in our field is a little wrapped in someone else’s engine: a shiny interface that’s usually powered by a large, general-purpose model running somewhere in the US. It looks impressive in the demo.
But selling is actually a very different proposition, and there are risks that sellers and firms are not fully aware of. When you start asking questions about who owns the data, who controls the costs and who is responsible when the output goes wrong, the glossy interface is pulled back to reveal something far less impressive and potentially dangerous.
The future of technology in the middle ground will be shaped by how seriously our industry takes those questions right now.
The four pillars
That’s exactly why, in April, we published the Mortgage Brain AI Charter. It lays out four pillars that govern how we build AI for this industry: cost, intellectual property, consistency and speed.
Take the cost. The current economy of third-party AI is unstable. Suppliers are running at a loss to gain market share, and analysts are predicting price increases of 10–30 times current prices as the market matures.
Consumers and lenders need to be completely transparent about how the data they share with these tools is used
Any technology company that builds its proposition on someone else’s model is running a business that cannot control its own prices. That cost will come somewhere and, with Consumer Duty already raising the bar on fair value, retailers can’t simply pass it on to customers.
AI in mortgages must be purpose-built and work with the infrastructure we have – small, low-cost, compact models designed for the job.
Then there is intellectual property. When client data flows into a third-party AI platform, it often leaves UK or EU jurisdiction and is processed in the US or elsewhere, where it can be used to train the next version of someone else’s model. Or worse, shared with hackers. This is a major problem in a regulated industry built on client trust.
The next phase of AI in this market will not be achieved by anyone who shouts the loudest about innovation
Consumers and lenders need to be completely clear about how the data they share with these tools is used, where it is processed, what barriers sit around it and what happens if something goes wrong. Moving sensitive client data through a public general purpose model creates an unimaginable risk.
Consistency is the third pillar, and this is where I think marketers will quickly lose patience with generic AI. Ask a social chatbot the same accessibility question twice and you might get two different answers. This is not acceptable for work that is subject to strict control procedures. The future of AI in our market lies in deterministic systems for parts of the process that require certainty, where AI is used surgically where judgment and language add real value. AI should support governance, not replace it.
And finally, speed. Speed is the easiest thing to market and the easiest thing to misuse. Not every step of the mortgage journey requires a general purpose model. Others need rules. Others need data. Others need someone. The skill is to know the difference, and that is only possible when you control the full stack.
AI has gone from a curiosity to something that every consumer company is encouraged to use
Where does this leave the mediator? It is in a stronger state than the sound suggests.
Consumers don’t need to be AI developers. They need to ask better questions of their technology partners. Whose model is this? Where does my client data go? Will the answer be the same tomorrow? What happens to my price if your supplier’s prices change?
The next phase of AI in this market will not be achieved by anyone who shouts loudly about innovation. It will be won by firms that treat AI as an infrastructure, carefully built, properly managed and accountable to the vendors and customers who rely on it.
That is the future of AI in the mortgage intermediary market.
Zahid Bilgrami is the CEO of Mortgage Brain
This article appeared in the May 2026 issue of Mortgage Strategy.
If you would like to subscribe to the monthly print or digital magazine, please click here.
