How Keystone is using AI to level up its broker proposition

As rewarding as working in the mortgage market can be, most brokers will admit their least favourite part of the job is the admin. That’s why the rise of Artificial Intelligence (AI) has caused such a reaction across the industry. With a few well-constructed prompts, AI can draft emails, produce reports or even generate code in seconds.

But it’s not a silver bullet. AI is only as good as the data it learns from and how thoughtfully it’s used. Get that wrong and it can slow things down, rather than speed them up.

At Keystone, we’re excited by what AI can do. But at the same time, we’re realistic about its limits.

Our AI journey began two years ago. After exploring various third-party systems, it became clear that none of them were quite right for what we wanted to do. So we decided to build our own solution.

The clearest opportunity for improvement was document handling. Previously, brokers uploaded files manually, selecting a category like “ID”, while case managers added another label such as “Passport” or “Driving Licence”. It worked but it was repetitive, time-consuming and far from efficient.

Using our archive of correctly labelled documents, we trained a machine learning model to recognise and categorise files automatically. Now brokers can upload and label 50 documents in the time it used to take to upload one.

Accuracy is already around 90%, and any low-confidence result is still reviewed by an experienced case handler to ensure total precision. The result is less admin, faster turnaround times and a much smoother journey for our broker partners.

For us, AI isn’t about replacing people. It’s about removing the tasks that slow them down, so they can focus on work that genuinely adds value.

AI will never be left to make lending decisions for a specialist lender like Keystone. But it’s already transforming the way we support brokers for the better.