AI now writes 85% of the code behind one UK bank’s business loans
Key Points
- AI agents write around 85% of the code in Shawbrook's digital business lending unit.
- 957 staff hold company-approved AI licences and have built over 350 custom tools.
- AI triages business loan enquiries and produces same-day indications of terms.
- Automated systems now handle routine savings enquiries and call summaries.
- The bank lists AI as its only emerging risk and has created an AI model risk committee.
Around 85% of the code in Shawbrook’s digital business lending arm is now written by AI agents rather than human developers.
The London-listed bank disclosed the figure in its interim financial report on Thursday (6 August), alongside a broader account of how far the technology has moved into its day-to-day operations.
Some 957 staff now hold company-approved AI licences, and colleagues have built more than 350 custom AI tools for their own productivity.
Shawbrook said it holds enterprise licences with two AI platforms and has issued updated acceptable use guidelines across the business.
AI business loans
The technology has reached the point where a loan application meets it before a person does.
In its small business division, Shawbrook has built a multi-agent system that triages incoming enquiries and produces an indication of terms, generating same-day terms for qualifying deals where the process previously took days.
A separate tool in development finance, named Origin, gives near-instant initial responses to developers seeking funding.
In commercial property, AI pulls transaction data, credit history, valuations and portfolio exposure into a single document for an underwriter to review, and produces annual portfolio reviews automatically.
The bank is targeting a 50% cut in the time spent on those underwriting tasks and a turnaround of less than a day on annual reviews.
Shawbrook said savings customers encounter it too. AI-powered messaging now handles routine enquiries, cutting manual processing across inbound messages by 18%, while automated call summaries and wrap-up notes have reduced post-call administration by around 20%. Contact volumes across calls and texts fell 35% year on year.
On the lending side, the bank has turned to AI for continuous monitoring of loans that may run into trouble, drawing on both structured and unstructured data to flag problem cases earlier.
AI also seen as a risk
In a somewhat ironic move, Shawbrook also lists AI as the single emerging risk on its risk register and has set up a dedicated AI model risk committee to oversee the more complex deployments.
It noted that during the period, the Bank of England, the Financial Conduct Authority and HM Treasury published a joint statement on the cyber resilience implications of frontier AI, warning of the growing speed, scale and sophistication of AI-enabled threats to the financial sector.
Shawbrook has since established a programme covering governance, vulnerability management and supply chain risk in response. Underwriters retain full accountability for lending decisions, according to the bank.
