Interest in AI among brokers is real, and it is rising. At the BIBA 2026 conference, brokers surveyed put AI and automation among their top priorities for the year — second only to attracting and keeping talent. The appetite is there. The problem is that interest isn't impact, and the two are easy to confuse.
Across the wider industry, the picture is sobering. On carrier-side numbers, only around 7% of insurers have genuinely scaled AI, with roughly two-thirds still stuck in piloting. That is a carrier figure, not a broker one, but it tells you how hard the last mile is even for firms with far deeper pockets. UK insurers have moved fast — more than half now run AI in at least some functions — so the direction of travel is not in doubt.
Independents face a particular bind. Most are small or micro firms, and cost is the barrier they cite most often. Broking is also a trust business: only about a third of UK insurance customers say they are comfortable with AI in the process, and usually only on the condition that a human stays involved. So the real question was never whether to use AI. It's how to capture the upside without losing the relationship — or falling foul of Consumer Duty.
1. Where do we start?
Start where the work is high-volume, repetitive and low-judgement: renewals, quote handling, mid-term adjustments, rekeying between systems, and the endless task of evidencing Consumer Duty. These are the places where time leaks out of the business every day without adding much that a client would notice or pay for.
For most brokers, the renewal report is the natural first move — it is, in effect, the broker's suitability report, produced again and again. In personal lines, the value tends to sit at the front end, in the speed of a quote. In commercial, it's about cutting rekeying so that experienced people spend their time on advice rather than admin. The honest verdict: start where the work is dull, frequent and low-risk, not where the technology looks most impressive.
In practice → put this decision on the one-page AI scorecard in the full guide.
2. Plug into our systems, or change how we work?
The advice to "redesign your workflows first" sounds sensible and is usually a trap. It is exactly where projects stall — months of process mapping before anything useful ships. Brokers live in Acturis, OpenGI, CDL and, above all, the inbox. A tool that ignores that reality asks people to work somewhere new, and they won't.
The better approach is AI that plugs into that spine through APIs and meets people where they already are. Embed first, evolve later. Let the tool prove itself inside the existing way of working, then reshape the process once you actually know what it changes. The honest verdict: fit the technology to the workflow, not the other way round.
In practice → put this decision on the one-page AI scorecard in the full guide.
3. Buy, partner, or build?
Building your own AI is a full department's worth of never-ending work — models to maintain, security to own, a roadmap that never ends. Almost no independent should attempt it. Buying or partnering is faster, cheaper and lower-risk, and it is the right answer for the overwhelming majority.
The real skill isn't building; it's choosing well in a crowded and noisy market, which is precisely where independent, non-conflicted advice earns its keep. There is a fourth option worth weighing honestly: switching on the AI features the incumbent software houses — Acturis, OpenGI, CDL — are adding to their platforms. It can be the path of least resistance, but it ties you more tightly to a single supplier. The honest verdict: don't build; buy or partner, and choose deliberately.
In practice → put this decision on the one-page AI scorecard in the full guide.
4. How much do we let AI decide?
In a regulated business that affects real people, AI should be assistive, not autonomous. It drafts, suggests and surfaces; it does not decide unsupervised. Every output should carry its source, a signal of how confident it is, and a named human to approve it before anything reaches a client.
Accountability cannot be delegated to a machine. The regulator is explicit on this, and Consumer Duty reinforces it: the firm remains answerable for outcomes regardless of what tool produced them. The honest verdict: keep AI on the assisting side of the line, always with a person accountable.
In practice → put this decision on the one-page AI scorecard in the full guide.
5. Will it work with our systems — and our people?
There are two problems here, and they are not equally hard. The technical one — connecting via APIs, handling unstructured data, doing it with the right security and permissions — is usually tractable. It is engineering, and engineering has answers.
The human one is harder. Adoption, not integration, is where most tools quietly die. If the tool comes into the inbox and management system people already use, it stands a chance. If it demands a new habit or a separate screen, it gets abandoned within weeks, however clever it is. The honest verdict: the technology will usually work; plan hardest for whether your people will use it.
In practice → put this decision on the one-page AI scorecard in the full guide.
6. Where do we keep a human in the loop?
Oversight shouldn't be left to chance or to whoever happens to be busy. Decide it on purpose, process by process, and write it down. Some things are non-negotiable: Consumer Duty outcomes, vulnerable customers, complaints, and anything that looks like advice all need a human firmly in the loop.
A sensible pattern is to start with heavy oversight and let AI earn its way down — from checking everything, to sampling, to reviewing only the exceptions — with the level set by risk, not by convenience. Name the subtler danger too: people either over-trusting the AI and waving things through, or under-trusting it and ignoring output that is genuinely useful. The honest verdict: design oversight deliberately, and revisit it as trust is earned.
In practice → put this decision on the one-page AI scorecard in the full guide.
7. Big bang or step by step?
Going all-in at once tends to create more problems than it solves — everything changes together, nothing is properly proven, and when something breaks you can't tell what. Start small instead. Prove one task works, then chain the next one on: intake, then triage, then recommendation.
As you go, the bottleneck moves — solve quoting speed and the pressure shifts to case handling, and so on. That is normal, and it is why this favours an ongoing advisory relationship over a one-off build: the work is a sequence of small, measured steps, not a single project with a finish line. The honest verdict: sequence it, prove each step, and let the bottleneck tell you what's next.
In practice → put this decision on the one-page AI scorecard in the full guide.
8. How do we know it's working?
Decide what success looks like before you start — otherwise you'll never be able to prove it afterwards, and every tool will feel vaguely worthwhile. Measure across four lenses: efficiency, client experience, staff experience, and the bottom line. Pick a few concrete measures under each, baseline them honestly, and review on a set cadence.
If a tool isn't moving a number that matters, stop — that is a result too, and a cheaper one than carrying on. The honest verdict: choose your measures up front, be honest about the baseline, and be willing to switch things off.
Capability is now competitiveness
For an independent, AI capability increasingly decides two things: whether you can stay competitive without having to sell, and — if you do sell — the multiple you command. It has moved from a curiosity to part of the firm's underlying value.
None of that needs a grand programme. One team, one task, clear measures, humans firmly in control, and an honest outside read on where to begin. Get the first decision right and the rest follow more easily than they look.
This article is general information, not legal or compliance advice. References to Consumer Duty and the regulator are for context only; check your own obligations with a suitably qualified adviser before relying on anything here.
Deciding where to start?
We help independent brokers pick the one task worth automating first, prove it works, and keep humans in control — without a grand programme. Start with the full guide, or talk it through.
Download the full guideSources: Survey of brokers at the BIBA 2026 conference (conducted by Mission, reported by Insurance Business UK); BCG "Build for the Future 2024 Global Study" (global carriers, industry context); Earnix "Insurance 2026: AI Trends Bulletin" and Lloyd's-market reporting via UK trade press; Guidewire "2026 European Insurance Consumer Survey" via UK trade press; BIBA written evidence to Parliament and BIBA 2026 materials. Figures verified June 2026; re-confirm before relying, as the picture is moving quickly.