Manufacturers say skip the new agents: process mining plus RPA saved 12,000 hours a year
shashib · x · 2026-10-07
At Infor Velocity Week, the manufacturing-floor question wasn't how to build agentic AI but where it belongs. Dr. Henning Dransfeld laid out a discipline many enterprise roadmaps skip:
- Diagnose first: use process mining to pinpoint where processes actually stall
- Automate deterministic work with RPA — Ring Container Technologies saved 12,000 hours annually just automating signed bills of lading
- Deploy agents only for decision support, where coordination and judgment are needed
The takeaway: mature ML models, process mining, and plain RPA are often the most reliable path to efficiency before reaching for a brand-new agent.
More from coding & agent
- A 3Blue1Brown-Style Explainer Rendered End-to-End in Rust by franken_manim — doodlestein · 2026-10-07
- Netlify AI Gateway Example: A Function That Returns Nano Banana-Generated Images — thisiskp_ · 2026-10-07
- Opus 5.5 builds 'black slopboxes' that beat human code on speed and quality despite ugly aesthetics — burny_tech · 2026-10-07
- Free official AI courses from Anthropic, Google, OpenAI and 7 more, in one list — nikola_mr64990 · 2026-10-07
- Theo's agent-built tsc-rs: $400k of Codex failed, $20k of Opus finished in 2 weeks — dejavucoder · 2026-10-07
- LLMs lack 'idempotency': rechecking their own large codebases yields contradictions — StephanSturges · 2026-10-07