AI training at an energy company shows domain expertise doesn't teach agent engineering
NumbersProtocol · reddit · 2026-09-16
Numbers Protocol shares lessons from an AI training engagement with an energy tech company, where employees proposed ideas like contract tracking, health-exam reminders, and approval histories.
One employee had a detailed contract-tracker concept—reading scanned agreements, distinguishing final versions from drafts, flagging renewal deadlines—but struggled with the prompting-and-debugging loop. What she actually asked for was a basic architecture, a data structure, and working examples she could modify.
Key takeaway: giving someone an AI agent still leaves them with software-structuring, troubleshooting, and verification decisions; knowing the job well doesn't prepare them for those parts. The proposed fix is starting the next session with a small working example around her contract process, with technical support available. The authors are upfront that this workflow is still proposed, not yet proven in production.
More from coding & agent
- Claude Code bypass permissions mode swaps built-in tools for Bash — stablequan · 2026-09-16
- Dev wishlist: coding harnesses should treat automations as one thing, not 100 chats — blixt · 2026-09-16
- Qwen 27B on a single RTX 5090 generates 8-scene animation from one prompt — Acceptable-Object390 · 2026-09-16
- 37-year-old non-coder builds DnD matchmaking site Tavernly with Claude Code — DopeDutchLemon · 2026-09-16
- ChatGPT's built-in Sites feature turns a single prompt into a live, publishable website — TawohAwa · 2026-09-16
- Blogger builds satellite-imaging financial due diligence tool on Doubao Seed 2.1 — karminski3 · 2026-09-16