GoFundMe exists to help people help each other. For Arnie Katz, Chief Product and Technology Officer, that mission has a direct operational translation: every week of cycle time cut, every backlog ticket cleared, every feature shipped faster helps more money flowing to the people and nonprofits who need it. With a team of 400 technologists and those stakes, velocity isn't a vanity metric, it's imperative.
From autocomplete to a factory of agents
GoFundMe was one of Augment Code's earliest customers, adopting the platform when LLMs were still new and autocomplete was the ceiling of what AI-assisted development could do. What started as engineers clicking through tab completions, with modest productivity gains, has transformed into GoFundMe running overnight multi-agent workflows spanning frontend, backend, performance, security, testing, and privacy and compliance.
Every morning engineers are presented with a queue of reviewed, ready-to-merge pull requests. PRD-to-pull-request cycles that once took months now take significantly less time. The code review agent has, in certain cases, reduced cycle time by one to two days.
"We are now way, way beyond AI just unlocking faster typing," Katz said. "The path to leverage is context, then intent, then automation."
GoFundMe's agent workflow
What that looks like in practice
Arnie described a meeting with a product manager, a designer, and one engineer. They recorded themselves talking through a new payment type for Giving Funds, GoFundMe's fee-free donor-advised fund product: the edge cases, the guardrails, the happy path. An hour of conversation. ChatGPT turned the recording into a spec. Augment took the spec and built the product, frontend, backend, feature flag, in another hour. Two hours of three people, for something that would have previously taken months.
That's the kind of outcome that changes how a team thinks about what's worth building.
GoFundMe's automation strategy
It also changed who does the building. A support team member noticed a recurring bug in PDF receipts because they were the one fielding the calls about it. They used Augment to trace the root cause, propose fix options, and open a pull request for an engineer to review.
A product manager, former finance background, built and shipped a feature during a 45-minute flight. They dictated what they wanted the product to be, turned that into a plan, broke the plan into tasks, and handed the tasks to Augment. When they landed, they walked engineers through the pull request. It went to production.
Expanding who can ship
"This is really expanding who can build, who can contribute successfully," he said. "It’s not only engineers who can do that anymore."
The backlog that never moved because it was never economical to prioritize is now clearing. The features that required two months of handoffs between product, design, and engineering are shipping in hours. The performance vulnerabilities that would have surfaced when a system went down are getting caught overnight, before anyone's paged.
For GoFundMe, the result is an engineering team increasingly focused on the calls that require judgment: architecture, product intent, and what to build next, while agents handle the execution. Arnie is already having daily conversations with engineers on his team whose roles have shifted away from hands-on coding over the past few months. He expects that number to grow.
"The goal is not that AI writes code," he said. "The goal is shipping outcomes faster while maintaining high safety, reliability, and trust."
For a platform moving billions of dollars toward causes people care about, that's not an engineering story. It's the mission.