OVERVIEW
Insurance
12,500 Employees
AI Adoption
Deploying agentic AI for claims intake and triage
"Adjusters heard 'agentic AI' and assumed we were automating their jobs away. We hadn't earned the right to skip that conversation. Once people saw the agent doing the triage grunt work and leaving the judgment calls to them, the tone in the room started to change."
Chief Information Officer
The Challenge
The company's claims process ran on manual triage: adjusters spent hours each week reading, sorting, and routing claims before they could do any actual adjusting. Backlogs grew every time volume spiked, and customers felt it in slower response times.
Leadership wanted to deploy agentic AI to auto-triage claims and pre-populate case data, but the small AI enablement team knew the real risk wasn't the technology. Adjusters across four regional hubs, with different systems and different levels of trust in automation, worried the agents were a step toward replacing them rather than supporting them.
The Approach
Being explicit from day one about what agents handles and what remains a human decision.
Embedding change champions inside each regional hub to surface concerns before they hardened into resistance.
Running structured feedback loops that fed directly into the agent's design.
Rolling out hub by hub, carrying lessons and fixes from each wave into the next.
Giving adjusters a direct channel to flag bad agent decisions and closing the loop within days.
Results & Impact
Average time from claim intake to adjuster assignment dropped 68%. Daily active use among adjusters reached 90% within four months, and the backlog that used to spike with volume stayed flat through two subsequent peak periods.
Sentiment shifted alongside the numbers: hubs that had a say in tuning the agent's behavior adopted twice as fast as hubs that only received a rollout announcement, confirming the necessity for trust-building work for future rollouts.
"We tracked trust the same way we tracked usage. Every hub had a different starting point, and the teams that got a say in how the agent behaved adopted it twice as fast as the ones that just got a rollout email."
Director of AI Enablement
Looking Ahead
The company is extending agentic AI into underwriting support and early fraud detection, using the same hub-by-hub trust model. A formal governance framework for agent decisions is next, alongside a company-wide standard for how much autonomy an agent earns as its track record builds.
