The AI Conversation Among UK Brokers Has Quietly Changed

Why are brokers suddenly talking openly about using AI day to day?

Two things happened roughly at the same time. AI tools became good enough at narrow, repetitive tasks — comparing wordings, drafting first-pass client emails, summarising long documents — that the productivity gain became hard to ignore. And enough brokers started using these tools quietly, on their own initiative, that admitting it stopped feeling like an admission of cutting corners.

The Bank of England and FCA's most recent joint survey on AI in UK financial services found 75% of firms already using AI in some form, with a further 10% planning adoption within three years. That's a sector moving well ahead of the UK business average — GOV.UK's 2026 AI Adoption Research puts overall UK business adoption at just 16%, with finance and information services among the few sectors significantly above that figure.

For a broker, the practical trigger is usually smaller than a strategic adoption decision. It's one renewal that used to take a full day taking three hours instead, told to a colleague over a coffee, repeated enough times that it stops sounding unusual.

What's actually driving brokers to trust AI more than they did a year ago?

Trust isn't really the right word for what's changed. It's closer to familiarity removing the stigma around admitting use. A broker running policy wording through Copilot to check for a missed exclusion isn't necessarily more confident in the tool's accuracy than they were a year ago. They're more comfortable saying out loud that they do it, because enough peers are doing the same thing that it no longer reads as a shortcut.

That distinction matters for how firms should respond. Comfort discussing AI use openly is good for normalising sensible practice and surfacing bad practice that used to stay hidden. It is not the same as verified reliability, and treating rising comfort as if it were rising trustworthiness is where the risk concentrates.

What happens when that comfort outpaces verification?

In January 2026, West Midlands Police admitted that an intelligence dossier used to justify banning football fans from a match referenced a fixture that never happened. Microsoft Copilot had invented it. The chief constable initially told a parliamentary committee that no AI tool had been used in preparing the material, then corrected himself weeks later. He retired shortly after the Home Secretary said she'd lost confidence in him, and an independent review found confirmation bias had shaped how the force used the AI-generated content in the first place.

Nobody in the review chain had checked the claim against a primary source before it shaped an operational decision affecting thousands of people. The hallucinated fixture read with exactly the same tone and confidence as everything else in the dossier, which is precisely why nobody flagged it.

That's the version of the broker scenario worth taking seriously. A policy summary drafted by AI that omits or misstates a clause doesn't look uncertain. It looks like every other summary a broker has read. The growing comfort with using these tools day to day is healthy. The absence of a habit for checking the output against the source document is where the real exposure sits.

Does the FCA expect brokers to have an AI policy?

There's no AI-specific rulebook, and the FCA has said directly it doesn't plan to introduce one. The regulator's published position is technology-neutral and outcomes-focused, relying on the Consumer Duty and the Senior Managers and Certification Regime to cover AI use the same way they cover everything else a firm does.

That doesn't mean AI use sits in a regulatory gap. It means the existing obligations already apply, whether or not a firm has written anything down. If an AI-assisted client communication is wrong in a way that affects the customer's understanding of their cover, that's a Consumer Duty question regardless of which tool produced the error. The House of Commons Treasury Committee's January 2026 report pushed back on the FCA's current approach, recommending the regulator publish clearer guidance on how existing rules apply to AI by the end of 2026 — but until that lands, firms are working from the frameworks that already exist.

A quick gut check: is your firm's AI use ahead of its verification habits?

The conversation has changed faster than most firms' processes have. A short, honest gut check this week:

• Which of those uses touch anything a client sees? Renewal reports, summaries, comparison documents, client-facing emails all sit in a different risk category than internal drafts.
  
  

• Is anyone checking AI output against the source document before it goes out? If the honest answer is "we trust it," that's the gap to close first.
  
  

• Could the firm reconstruct how a specific piece of client work was produced if asked? Not for every draft, but for anything that informed a recommendation. This is what the West Midlands Police case exposed when nobody could trace where the fabricated claim had come from until journalists and a parliamentary inquiry forced the question.
  
  None of this requires slowing down the adoption that's already happening. It requires the verification habit to catch up to it.

The conversation isn't over. It's just changed shape

Brokers aren't more trusting of AI than they were a year ago. They're more comfortable admitting they use it, which is a different thing entirely, and a healthier starting point than pretending the tools aren't already in daily use across the market. The firms that come out ahead won't be the ones that adopted fastest or the ones that stayed most cautious. They'll be the ones whose verification habits kept pace with how normal AI use became.