The Bottleneck in Lending Isn't Underwriting Anymore. It's What You Do with the Freed Capacity.

AI reads a bank statement in minutes. Income, commitments, HMRC activity, unpaid items — extracted and categorised before an underwriter has opened the file. The first-pass work that used to take an hour per case is largely done.

So where's all the extra lending?

Mostly nowhere. And I think the reason is the same one playing out across software right now: the bottleneck has moved from the work itself to the organisation around it.

Lending Culture Was Built for the Old Constraint

Underwriting was slow and expensive, so everything was optimised around protecting underwriter time. Queue the deals. Triage hard. Decline anything that costs more to assess than it's worth. Plan capacity quarters ahead.

Then AI removes the constraint, and the default response from most lenders has been: keep the exact same process, just run it faster. Same credit box, same deal types, same referral thresholds — quicker turnaround.

That's not nothing. Faster decisions win deals. But it leaves most of the new capacity on the table. Work planned for the quarter is done in a month, and nobody planned what comes next. So the surplus gets absorbed by internal noise — or worse, someone concludes half the credit team is now superfluous and cuts it.

That's the wrong lesson. A smaller team writing the same book isn't an edge. Anyone can do that. The edge is the same team writing more.

What Does Surplus Underwriter Capacity Actually Buy a Lender?

• Deals you used to decline on cost, not risk. The thin-file case that needed two hours of statement work to assess properly was never worth it at the old cost per decision. It is now.
• Referrals worked properly instead of rubber-stamped or bounced. When the mechanical work is done for you, the judgement work gets the time it deserves.
• Back-book monitoring, not just front-door checks. Most lenders only look hard at a customer once — at application. Continuous statement review of the existing book was always desirable and never affordable.
• New segments. The adjacent market you didn't enter because you couldn't staff the underwriting for it.

None of this happens by accident. It's a leadership decision, not a technology one. The technology is done — it works, it's deployed, it's cheap relative to what it replaces. What's missing in most lending organisations is the mindset shift: from “how do we do the same with fewer people” to “how do we lead the same people to do more.”

The winners in the next cycle won't be the lenders with the fewest underwriters. They'll be the ones who worked out, first, what to do with the time AI gave back.