Project Summary

When someone applies for a loan, the answer is often no. That leaves the applicant no better off and no wiser about what would need to change, and it can push them towards far more expensive lenders. Credit unions want to say yes more often, and responsibly, but they are small organisations with little room to experiment on real people’s finances. 

This project builds a computer simulation that lets them experiment safely. Working with incuto, a technology provider to credit unions across the UK, we will use patterns in anonymised past lending records to create realistic “member personas”. Each is represented by an AI agent that behaves as that kind of member might over time, responding to loan offers, repayment plans, advice and events such as a rise in energy bills or a drop in income. 

Credit unions can then test questions they cannot safely test on real members: would a smaller loan have been repaid? What advice turns a no today into a yes later? The simulation can also assess how members might engage with payroll-linked savings schemes, helping credit unions design products that prevent people needing to borrow in the first place.

Partner Organisation