Roy Feig | Honeycomb Insurance | How Do You Build an Efficient Insurance Product?

In this episode

• When an entire industry underwrites categorically and exits whole regions, there's an opening for whoever can assess risk at the level of a single property. Honeycomb built its IP exactly there: spotting the good buildings everyone else gives up on.
• The automation was built in stages: first understand what underwriters look at and give them rules, then a model that predicts their decision, and on top of it an AI layer that decides on its own. That took them from zero automation to about 60% of requests, with a target of 80%.
• Every product manager owns one persona and one metric. The agent PM is measured on retention of agencies that come back and produce policies every month, not on closing deals.
• Even when the system declines, you can still deliver value: explain to the agent why the property was declined, what will work in the next deal, and help them explain to their client what to do. That keeps the experience positive without a salesperson in the loop.
• Not every automation should reach 100%. Honeycomb samples part of every model's decisions and routes them to human underwriters, to measure accuracy and correct the model over time.
• Anyone moving into a new industry should listen to the experts, be curious, and ask why. Some of the best things Roy did were things he chose not to build, just because he asked why the industry does it that way.