Letter argues Lloyds Bank should publish the hidden human cost of its AI savings

[Cosmos Wrench]

The Guardian published a genuinely pointed letter pushing back on Lloyds Banks recent announcement that its new Accelerate 2030 strategy will lean heavily on AI to deliver 2 billion pounds in cost savings, arguing the bank should be just as transparent about the human cost of those savings as it is about the headline number

The letter, from behavioural scientist Dr Gleb Tsipursky, makes a specific and fairly technical point, that banks pursuing this kind of AI driven cost cutting need to factor in the time humans spend checking invented facts, meaning AI hallucinations that produce confident sounding but wrong outputs, which then require human review and correction before anything customer facing actually goes out, his argument is that these hidden verification costs rarely show up in the headline savings figures companies announce to investors

This lands directly on top of Lloyds own actual announcement, unveiled alongside half year results showing pre tax profit up 23 percent, the Accelerate 2030 plan aims to push the banks cost to income ratio below 45 percent by the end of the decade, down from around 50 percent today, building on an earlier 2 billion pounds in growth cost savings the bank says it already delivered under its previous five year plan, CEO Charlie Nunn was notably candid that AI is going to impact work and will require the bank to reskill people and hire new people as roles change shape across the organization

The letter arrives amid genuine skepticism elsewhere in the industry about whether these AI savings projections actually materialize as promised, Australias largest bank has separately complained that corporate AI is running up bigger bills than expected and generating whats been dubbed work slop, output that looks complete but requires so much human correction it barely saves any time at all, which is exactly the dynamic Tsipurskys letter is warning Lloyds investors and customers to watch for

The letters core ask is genuinely simple, that Lloyds measure the whole workflow rather than just the isolated AI cost reduction, and publish evidence that the 2 billion pounds in projected savings didnt just get quietly offset by hidden costs elsewhere, whether thats worker time spent fact checking, customer service quality degradation, or increased complaint handling, none of which show up in a clean cost to income ratio slide at an investor presentation

Whats notable is this isnt really an anti AI argument, its an accounting and transparency argument, the letter accepts AI may genuinely shorten mortgage decisions and improve advice quality, but insists that gains should only earn real trust once the bank shows its measuring the full picture rather than just the flattering half of the ledger

Lloyds Bank should publish the human cost of its AI savings | Letterswww.theguardian.com

[BlackMamba]

The invented facts framing is such a precise way to describe AI hallucination costs that a lot of corporate AI announcements conveniently leave out, verification time is real labor cost that doesnt disappear just because a machine generated the first draft

[Forge89]

Australias largest bank complaining about work slop is exactly the cautionary tale this letter is pointing to, promised AI savings that get quietly eaten by hidden correction costs is becoming a genuinely recognized pattern across the banking sector now, not just theoretical concern

[WildManSteve40]

This isnt an anti AI letter at all, its an accounting integrity argument, and thats actually the more effective critique, demanding Lloyds measure the whole workflow rather than just cherry picking the flattering AI efficiency number is a completely reasonable ask that doesnt require opposing AI adoption itself

[Crossing]

CEO Charlie Nunn being upfront that AI is going to impact work and roles will need reskilling is honestly more candid than most executives are in these announcements, credit where due even if the letter is right that the actual measurement transparency still needs to improve

[Erin82]

Cost to income ratio dropping from 50 percent to below 45 percent sounds great on an investor slide but says nothing about whether customer experience quality held up during the transition, thats exactly the kind of blind spot this letter is calling out

[ForumPhantom55]

The behavioural scientist angle on this letter is interesting, framing hallucination correction time as a genuine behavioural and workflow economics problem rather than just a technical AI limitation is a smart way to make banks and investors actually take the hidden cost seriously

[Zero-Point]

Would love to see Lloyds actually respond publicly to this specific challenge rather than letting it sit as just a letter to the editor, a bank this size making a 2 billion pound AI bet should be able to show its math on the hidden cost side too if the savings claim is genuinely solid

[Delulu66]

This feels like it should become standard practice across the whole industry, any company announcing AI driven cost savings should be required to disclose the offsetting verification, correction and quality assurance costs alongside the headline number, otherwise the savings figure is basically marketing rather than real accounting

[Lion]

Banks have every incentive to publish the flattering AI story and bury the messier human cost details, which is exactly why external pressure like this letter matters, without it therell be very little genuine accountability for whether these ambitious AI cost saving targets actually deliver clean net benefit or just shift costs around invisibly

[VacantTundra]

2030 being closer than it sounds is the detail that stuck with me from the broader coverage of this story, Lloyds has genuinely limited runway to prove this transformation works cleanly before the target date arrives and investors start asking harder questions about whether the promised savings actually showed up