Scout

Zapier has maintained more than 8,000 integrations over 14 years, and the third-party APIs behind them keep changing. Lisa Orr, an engineering leader there, calls the result app erosion: tickets arrive faster than they are resolved. Scout was Zapier's second experiment against that backlog, following a first one that moved support staff from triage to fixing bugs directly, scoped to four target apps under mandatory engineering review.

Orr's team built Scout after watching where the work actually went, and found that half the time was spent assembling the context needed to understand a problem. Scout has three parts. Context Analyzer gathers tickets, error logs, API documentation and source code. Diff Generator writes the correction from that context. Test Generator produces tests to validate it. Orr reports a 97% merge request success rate, with 3% needing minor edits, and resolution time falling from 72 hours to a few minutes.

The same name appears in a second place. In Shopify's Q3 2025 results, president Harley Finkelstein describes an internal tool called Scout that analyses merchant feedback to inform product decisions. Nothing in either account connects the two beyond the name.

What links the entries is a shape of tooling: generative assistance pointed at the maintenance load that accumulates around a large integration surface. Orr treats the change as organisational, with the line between support and engineering fading. Whether a 97% acceptance rate survives outside Zapier's own catalogue goes unexamined.