Copy article

Mortgages in minutes? Why ING’s move to "autonomous" AI needs a seatbelt, not just an accelerator

ended 11. December 2025

The headline promise of cutting mortgage approvals from "months to minutes" is enticing, but we must look at the cautionary tales coming out of the insurance sector before we cheer.

​We have already seen what happens when finance automates too fast. Major US insurers like Cigna and UnitedHealthcare are currently facing class-action lawsuits for allegedly using algorithms to deny claims in an average of 1.2 seconds, often overriding human doctors (Source: https://www.cbsnews.com/news/health-insurance-humana-united-health-ai-algorithm/).

​Closer to home, the risk of "black box" bias is just as real. Citizens Advice has repeatedly warned of an "ethnicity penalty" in car insurance, where algorithmic pricing charges people of colour hundreds of pounds more based on data proxies rather than actual risk (Source: https://www.citizensadvice.org.uk/about-us/media-centre/press-releases/citizens-advice-sounds-the-alarm-on-280-car-insurance-ethnicity-penalty/).

​This is why ING’s restraint (Source: https://www.thebanker.com/content/a9bd62ae-bd4d-4c21-a1f4-3a4282e58efb) is the real story here. By vetting this technology against 140 separate safety checks, they are acknowledging that "neutral" data is rarely neutral. If we feed historical lending data into an autonomous agent without guardrails, we don't get efficiency; we get automated discrimination at scale.

​​Bahadir Yilmaz is right: a computer that does everything is a disaster waiting to happen. The goal of automation in finance shouldn't be to remove human staff, but to liberate them from data-drudgery so they can actually advise people on the biggest financial decision of their lives. If we hand over lending criteria entirely to machines just to save a few quid on headcount, we aren't just risking errors; we are inviting systemic bias into the housing market.

​We need to stop asking if AI can do the job, and start asking who takes the blame when the machine gets it wrong.

​We’d like your views:

  • ​The "1.2 Second" Risk: With insurers being sued for "instant denials," if a bank rejects a mortgage in minutes, can they legally prove they actually reviewed the file?
  • ​Invisible Bias: How do we stop "Agentic AI" from redlining by postcode or ethnicity, using "neutral" data points to unfairly penalize vulnerable groups?
  • ​Liability: When an autonomous agent denies a loan based on a "black box" calculation, who is liable for the discrimination? The bank, or the AI vendor?
  • ​Computer Says No: How do we ensure customers get a human explanation for a rejection, rather than a generic "computer says no" derived from complex data modelling?
  • ​The Human Role: ING suggests this won't kill jobs, just change them. Is that realistic, or is it corporate spin to hide upcoming redundancies?

4 responses from the Newspage community

Copy all

Star Quote
Copy

Rohit Parmar-MistryVerified
Founder at Pattrn Data
We are danger of sleepwalking into a real-life Little Britain sketch. The promise of 'mortgages in minutes' sounds brilliant, right up until you get a rejection that nobody, not even the bank manager, can actually explain.

Credit where it’s due: ING is trying to put a seatbelt on this. They’ve reportedly implemented 140 safety checks, which is far more diligence than we usually see. But guardrails in a boardroom presentation are very different from the real world. Only time will tell if those safety nets actually hold up when thousands of messy, complex human applications hit the system at speed.

If they don't, we create a world where a customer asks why they’ve been denied, and the advisor just shrugs at a server rack. That isn't progress; it's just 'Computer Says No' with a higher electricity bill.

Use AI to kill the paperwork, absolutely. But unless you want to automate disappointment, you must keep a human in the loop to sanity-check the logic.
Copy

Editor
Mitali DeypurkaysthaVerified
AI Governance Strategist & CMO at AISensa
Cutting mortgage approvals from months to minutes is exciting, but only if we don’t turn the housing market into a two-tier highway. AI is a car: powerful, fast, game-changing. But it runs on fuel, and our fuel is data. If that data is biased, the engine doesn’t just misfire. It drives some people off the road entirely.

We’ve already seen the warning lights. Citizens Advice exposed an “ethnicity penalty” in car insurance. The UK’s own NPL found facial-recognition systems misidentify Black faces 5.5% of the time and Asian faces 4%, compared to 0.04% for white faces. That’s not a glitch; that’s a mirror.

If we rush mortgage automation without guardrails, we risk a future where some buyers get keys in minutes while others hit invisible walls. AI should lift all boats, not widen the wealth gap at machine speed.
Copy

Patricia McGirrVerified
Founder at Repossession Rescue Network
Mortgages aren’t takeaway orders, and turning them into one is how you end up with a misselling scandal to rival PPI. ING at least recognise that speed needs guardrails. Their checks may help win market share, but they also keep people first. The real danger is lenders chasing minutes to approval without asking what gets lost in the rush. A mortgage is a twenty five year commitment. It must be measured, explained and right first time. Without that, customers end up shouting at the ghost in the machine when the algorithm says no. Tech should support good decisions, not see them as a transaction where speed is the key metric.
Copy

Colette MasonVerified
AI Ethics Consultant at Clever Clogs AI
The concern here is that AI doesn't create the same output twice when the task is complicated. Spin the "review the customer circumstances" wheel twice with the same mortgage application and get different answers? How does a bank prove in court they made the right decision and not the biased or inappropriate one? Automating aspects of the soul-destroying wall of paperwork brokers face is brilliant. Automating the actual lending decision when the system can't show its working and might be baking in postcode discrimination? That's liability roulette. Brokers are duty bound to show the recommendation is fit for the borrowers' needs. ING's 140 safety checks exist because they understand the difference between using AI to organise documents and trusting it to decide who deserves a home based on algorithmic guesswork that changes every time you pull the lever.