Artificial intelligence is moving beyond experimentation in the delegated authority market as insurers and managing general agents (MGAs) shift their focus from testing the technology to deploying it at scale.
Technology leaders told Insurance Business the conversation has moved away from whether AI can deliver value and towards the practical challenges of embedding it into day-to-day operations, with data quality, governance and operational readiness emerging as the biggest obstacles.
Graeme Asquith, UK managing director at mea platform, said firms are increasingly asking practical questions about implementation rather than exploring AI as a standalone productivity tool.
"We've definitely seen a shift to those more practical operational type questions," he said, describing a move away from AI being bolted onto existing systems towards agentic AI operating natively within business workflows.
Rather than replacing people, Asquith said AI's greatest value currently lies in automating repetitive operational processes. Even so, he stressed that digital workforces still require governance and human oversight rather than being left to operate independently.
Gavin Lillywhite, SVP business development for Xceedance's UKI and Europe region, said the most immediate returns are appearing in back-office functions, where firms are replacing manual processes with agentic AI.
He estimated efficiency gains of between 30% and 50%, although he cautioned that results depend heavily on data quality and how AI models are deployed.
"AI doesn't come for free," he said, pointing to token costs and the risk of "compromised outcomes" where poor-quality data undermines otherwise capable technology. He believes underwriting will take longer to see comparable gains.
"I think we're probably at the tip of the iceberg," Lillywhite said, arguing that operational efficiencies will become visible well before AI materially improves underwriting judgement.
Peter Bammodu and Andrew How, directors of insurance at Earnix, said AI is also creating new opportunities at the front end of the underwriting process. How said AI can rapidly ingest broker submissions that might otherwise remain unopened for days, dramatically reducing response times, while Bammodu said efficiency and competitive advantage were the themes dominating conversations across the market.
While AI capabilities continue to improve, several interviewees argued that the industry's underlying data remains the biggest constraint on adoption.
Charles Rowley, founder of DA Strategy, said delegated authority has wrestled with data quality for decades, recalling industry discussions around real-time messaging between MGAs and carriers more than 20 years ago.
Referring to the absence of a universal standard for data exchange, Rowley said: "It's still not."
He argued that widespread reliance on bordereaux and spreadsheets continues to force firms to re-key information between systems instead of allowing data to move automatically.
The issue extends beyond the delegated authority market. The CI&T and Reuters report Strategising for the AI Insurance Revolution found that underwriters see data quality as the single biggest barrier to AI adoption, rather than the capability of the technology itself.
Several interviewees also drew clear boundaries around where AI should and should not be used.
Rowley argued claims handling should remain in the hands of "a knowledgeable, well trained, experienced human being", warning that AI struggles where outcomes depend on experience, judgement and multiple possible decision paths.
Asquith made a similar point about brokers. As administrative work becomes increasingly automated, he expects brokers' value to become even more concentrated around advice, advocacy and structuring complex risks, the areas of the profession AI "won't take away."
That reflects a broader shift across the MGA market. Rather than replacing underwriters or brokers, firms are increasingly using AI to remove repetitive manual work, improve operational efficiency and give experienced professionals more time to focus on higher-value decisions, a shift already visible in newer MGAs building AI into underwriting from the ground up.
