The boring work behind successful AI agents: data and process cleanup decide outcomes
bigdata · x · 2026-11-17
Ben Lorica's Gradient Flow research on agent deployments in banks, utilities, manufacturers and retailers finds that getting agents into production tests your existing data and systems as much as your AI capabilities. Key points: limit what agents can do rather than how they reason (Duke Energy uses agents for grid-planning simulations with human review; DBS Bank chains 70-80 specialized agents for credit memos); Lyft warns overly scripted workflows break on unforeseen situations and routes fraud/safety issues to humans; previously distilled rules mostly hold in production; and published numbers deserve skepticism since most come from vendors or the companies themselves. The core argument: years of data cleanup and process documentation are the real head start.
More from coding & agent
- A practical CI prompt: make your tests and builds use every CPU core on the runner — DanielLockyer · 2026-11-17
- J.Doe enters private beta: proactive decision cards instead of empty prompt boxes — umesh_ai · 2026-11-17
- Building an Obsidian music notation library and practice plugin with Fable 5.2 — dSebastien · 2026-11-17
- Mistral makes Vibe free in Europe for 3 months, claims coding edge over GLM and DeepSeek — qtnx_ · 2026-11-17
- Cheap tokens, expensive agent: Flash hits $11.97/task vs Luna's $0.05 in Databricks AI Index — krishnan · 2026-11-17
- HADAL: a Mariana Trench dive game fully generated by Fable 5.2 in one 373KB HTML file — AravSrinivas · 2026-11-17