Disputes are a painful process for service pros. Hard-earned money they worked for gets disputed by customers for all kinds of reasons, and once Housecall Pro flags that a dispute has started, the pro is on their own to manually fight it through the platform. This is a slow, tedious task pulling them away from paying jobs in the field.
The numbers made the cost obvious: dispute win rate hovered around 20%, and only 49% of disputes were even submitted. That's a lot of money left on the table.
We built an agentic experience that removes the manual work entirely: since job details already live in Housecall Pro, a dispute agent gathers the evidence itself. It pulls customer communication, before-and-after photos, estimates, and invoices, and compiles all of it into a package.
The open question was what happens when a pro doesn't act before the submission deadline. I pushed to auto-submit the AI-generated package by default — more submissions at the same win rate means more disputes won and more money pros keep — but that meant trusting AI to act on a pro's behalf in something sensitive, before they'd had a chance to build trust in it.
The team decided to turn auto-submission on by default, disclose it clearly to pros, and let anyone who didn't want AI acting on their behalf opt out. We rolled it out in phases: a view-only preview that auto-submits as-is, then the ability to add supporting content, then generative AI tools to enhance the package further.
An AI-enabled dispute evidence feature that gives pros the highest likelihood of winning their disputes, with no manual work required. Pros can edit and enhance the package with AI's help, or simply let it auto-submit ahead of the deadline. Dispute win rate jumped from 20% to 29%, and submission rate jumped from 49% to 100%.
What pros don't see is the work happening in the background: the agent learns from disputes it has won and lost, training itself to keep improving future odds. If a dispute reason is "product not received," a simple before-and-after photo proving delivery may be enough to win it — the agent picks up on patterns like this and adjusts its packages accordingly.
