AI scheduling that groups similar parts and picks the best machine

Added 2026-10-14

Role

scheduler, operations manager

The problem

In a low-volume, high-mix job shop every order differs. The ERP produces a schedule that meets due dates and material timelines but ignores efficiency: it does not group similar parts to cut setups or send each part to the machine best suited to it. The experienced schedulers know how to do this, but that know-how lives in their heads.

Step by step

Interview schedulers and programmers: how do they group jobs, which machine do they prefer for which parts, and why?
Write those rules down (families, setups, machine capabilities) and check them against past schedules.
Extract part data: material, size, features, operations; this is what the geometry matching uses.
Build or configure a scheduler that respects due dates and applies the grouping and machine rules.
Compare its schedule with the human one for a few weeks before switching.
Let schedulers adjust and record why, to improve the rules.
Track on-time delivery, setups per week and machine utilisation.

Tools

Custom algorithm built in-house with Rotman School of Management students through the World Trade Centre Toronto Growth Development Program
Commercial alternatives: ProShop, Fulcrum, PlanetTogether, Siemens Opcenter APS
Python with an open-source optimiser (Google OR-Tools) for a do-it-yourself pilot

Result

J/E Bearing & Machine Ltd. (precision machining job shop, Tillsonburg, Ontario), built with Rotman MBA students in the Toronto Region Board of Trade's Growth Development Program: on-time delivery improved by 37%. A related quality project cut a two-hour data-entry task to under a minute (Toronto Region Board of Trade, May 2026).

As reported by the source; most figures are vendor or self-reported and not independently verified.

Human review: cautions

Single published figure, from a program promoting its own course.
Capturing the schedulers' knowledge is most of the work; involve them from the start.
Keep due-date commitments as hard constraints; efficiency gains must not hide late orders.

Try it yourself

You are a job-shop scheduler. I will paste jobs (job, part, material, size, operations, due date) and machines (capabilities, size limits, current queue).
1. Group jobs into families that share setups.
2. Assign each job to the best-suited machine and explain why.
3. Propose a sequence for each machine for the next week that meets due dates and minimises setups.
4. Flag jobs that cannot meet their due date.
Show assumptions.

Want more like this? Join the newsletter

Sources

https://bot.com/News/From-Gut-Feel-to-Algorithm-How-One-Manufacturer-Used-AI-to-Rethink-How-Work-Gets-Done