Insurance is running out of room to pretend artificial intelligence is still a side project. For years, the industry has treated AI as a useful but distant prospect: good for pilots, productivity demos and conference panels but not yet central to how business is written, serviced and claimed. That position is becoming harder to defend. Even in July last year, McKinsey was advising that insurers need to stop “dabbling” and instead “fundamentally rewire how they operate” across underwriting, claims, distribution and customer service. The industry is at the beginning of its AI revolution, but the starter’s gun sounded months ago.
“We really are more at the start of it than at the end,” said Andrew Broughton (pictured) CEO Network Insurance House.
Broughton suggested that the industry could still be underestimating how big the operational shift could be. Broughton leads one of the largest equity brokers in the Steadfast Group, placing about $900 million in GWP and generating just over $100 million in fee and commission income.
The first phase of AI in insurance has largely been about experimentation. The next one looks more like industrialisation: faster data extraction, quicker servicing, smarter claims communication, tighter platform connectivity and more automation across standardised parts of the book. The important point for brokers could be that AI does not need to eliminate them to materially change broking economics. It only needs to compress the time and labour tied up in the administrative middle of the business.
Broughton put the scale of that shift in stark commercial terms:
“This whole AI play, this technology play, this platforms play, the operational efficiency play, the claims interaction, the direct connectivity – this all effectively impacts what is 94% of our book,” he said.
This is not a niche use case – it’s nearly the whole machine – and it fits with what the wider market is saying. The National Insurance Brokers Association (NIBA) report released late last year, Ready or Reacting? Shaping the Future of the Insurance Broking Profession, found that 83% of respondents expect technology and automation to have a significant impact on broking by 2035, but only 61% feel prepared for it. The implication is not subtle: the disruption is widely expected, but the industry is still not organised for it.
That’s despite insurance being full of work that AI is well suited to attack first including submission handling, product comparison, file summaries and claims updates. These are some of the parts of broking built around repetition, standardisation and high volume that now sit squarely in AI’s strike zone. For larger brokers in particular, the upside is obvious: lower servicing friction, faster response times and better use of skilled people.
That could be where the more alarmist reading of AI’s impact starts to break down. If AI is coming hardest for administration, that does not mean it is equally suited to the moments when clients most need a broker.
“The thing that’s coming out is that the role of advice is always going to be there,” said Broughton.
There is mounting evidence he is right. Vero’s 2026 SME Insurance Index found businesses were relatively comfortable with AI helping with quote pre-fill, product comparison and round-the-clock policy queries but support dropped sharply when AI was framed as making judgments on claims. The same research said the findings reinforced the importance of brokers keeping expertise and trusted relationships front and centre, particularly at claim time.
Even in the days of the Lloyd’s coffee shop, the strongest broker value was never in clerical transfer, it was in interpretation, choice, advocacy and reassurance when a client has a difficult claim, an ambiguous wording problem or a real commercial loss unfolding in front of them. In those moments, the client does not want clever automation alone. They want a person who understands the cover, can speak to the insurer and can take control.
The hardest, most broker-relevant truth in AI may be that its impact on insurance probably is about to explode, but unevenly. The administrative core looks exposed. Advice and claims advocacy looks safer. The brokers who use AI to strip out friction and refocus on those higher-value functions may end up stronger, not weaker.
That is the real divide opening in broking now. AI may not replace the adviser but it could ruthlessly expose everyone who was never doing much more than processing in the first place.
