Why Document Crunch Chose Bagel AI Over Building It In-House, and Scaled Its Discovery Process into One Engine Wired to Claude

Founded 2019 · Atlanta · Part of Trimble

Document Crunch’s product team already lived inside AI tools before they ever talked to Bagel. Every PM ran their own prompts through Claude or ChatGPT, digging through customer calls and Slack messages for patterns. Senior Product Manager Marcus Erickson still went looking for something more, and the question he had to answer first was whether his team needed to buy anything at all, or could just build it themselves.

The build-vs-buy question that comes up every day

Document Crunch reviews construction contracts and project documents with AI, catching conflicts before they turn into disputes. The product team runs eight-week sprints full of customer calls, Slack requests, and Momentum recordings. All that raw material already existed. Getting one consistent read out of it across five or six PMs did not.

Marcus looked at 10 to 15 tools before he picked Bagel.

“I looked at about 10 or 15 different products, and I really thought that Bagel had the best ability to integrate across a whole bunch of different streams like Slack, Momentum, Salesforce, bring all that together.”

Claude was the harder comparison, since his team already used it every day for exactly this kind of pattern-finding. Marcus gives colleagues the same answer whenever they ask why Document Crunch pays for Bagel instead of building it in-house: Bagel runs one engine that processes every signal the same way, every time, with real work behind how the ideas get ordered. Building that internally meant redoing work Bagel had already done, instead of spending that time on the product itself.

The feature that was never on the roadmap

Customers kept mentioning one request, call after call: the ability to write their own prompts to review contracts, instead of routing everything through Document Crunch’s CS team. Nobody had put it on the roadmap.

“That was never on our roadmap, people talked about it, but really let us see this has been asked for by 70% of our customers and prospects. So it let us drive and change our roadmap to make sure we fit that in earlier.”

Then the MCP put Bagel’s engine inside Claude

Marcus writes his own Claude skills and works in Claude Cowork every day, so once Bagel’s MCP went live, he plugged it straight into that setup. Bagel handles the volume, turning every call and ticket into evidence Claude alone couldn’t process. Claude does the reasoning on top of it, joining that evidence with the team’s own product data to answer questions a dashboard never could.

The tool built for evidence and the tool built for building now talk to each other automatically.