Systems | Development | Analytics | API | Testing

Claude has become the place where a lot of everyday work gets done, and analysis and reporting are no exception. In Databox’s recent survey, Using AI You Don’t Trust, 86% of business users said they reach for an AI tool like Claude or ChatGPT before any other tool for analytical work. Honestly, the pull toward these tools is easy to understand. Feed Claude a spreadsheet, and it can synthesize your data, spot trends, and turn raw numbers into written explanations and reports in minutes.

An AI that drafts a listing or summarizes a lease is now ordinary in real estate software. The change worth watching is software that can carry out work inside the systems a business actually runs on. That change is easy to underestimate. In a demo, an agent that schedules a viewing or processes a renewal can look straightforward.

Jared Lynskey recently wrote a sponsored post, trying out Patch with his app, maru. This article is a follow up about using it in production on Fly.io, including any problems he encountered. Last month I moved maru’s over-the-air updates from my own expo-updates server to Codemagic Patch. Everything in that post ran against Patch’s local evaluation stack on my laptop. The section on production was a plan, with a note to myself to replace it once it had actually happened.

Traditional software monitoring builds reliability by tracking uptime, latency, and error rates. As artificial intelligence moves into core enterprise operations, establishing robust observability in AI becomes essential to measure what systems do once live, moving far beyond basic prediction accuracy and server health. Modern non-deterministic models and autonomous AI agents act rather than just predict.