Adding people or machines changes nothing unless they go to the bottleneck, the place that holds everything up. Here are four examples that an AI simulated in earlier articles, showing what helped where: picking and packing, trucks waiting, inspection and parts supply. The AI used the tool it calls over MCP (sweep_process_parameters), and each figure is the mean of 20 runs. A water spider is the worker who brings parts to the assembly line.
1. What was added, and what came of it
| Example | Set-up | What is measured | Result |
|---|---|---|---|
| A warehouse's afternoon shipping (a 15:45 cutoff) | Four pickers, two packing stations | Orders missing the cutoff | 25.2 orders |
| A warehouse's afternoon shipping (a 15:45 cutoff) | Four pickers, three packing stations | Orders missing the cutoff | 25.2 orders |
| A warehouse's afternoon shipping (a 15:45 cutoff) | Six pickers, three packing stations | Orders missing the cutoff | 1.1 orders |
| A distribution centre's morning trucks | One person at reception, eight docks | Time from arrival to unloading | 23.6 minutes |
| A distribution centre's morning trucks | Two people at reception, eight docks | Time from arrival to unloading | 8.3 minutes |
| A distribution centre's morning trucks | Two people at reception, ten docks | Time from arrival to unloading | 8.2 minutes |
| An assembly plant's first hour (11 orders) | A 3-minute test, four AMRs | Orders shipped | 10.8 orders |
| An assembly plant's first hour (11 orders) | A 3-minute test, six AMRs | Orders shipped | 10.8 orders |
| An assembly plant's first hour (11 orders) | A 2-minute test, four AMRs | Orders shipped | 11.0 orders |
| An assembly line (takt 60 seconds) | One water spider, two kanban cards | Line stopped for lack of parts (total) | 60.8 minutes |
| An assembly line (takt 60 seconds) | One water spider, three kanban cards | Line stopped for lack of parts (total) | 36.6 minutes |
| An assembly line (takt 60 seconds) | Two water spiders, two kanban cards | Line stopped for lack of parts (total) | 0.0 minutes |
2. What they show
- Adding away from the bottleneck changes nothing. With four pickers, the warehouse still misses 25.2 orders with three packing stations; with two people at reception, the distribution centre's trucks still take 8.2 minutes with ten docks; and with a 3-minute test, the plant still ships 10.8 orders with six AMRs.
- Adding at the bottleneck changes a lot. A second person at reception cuts the trucks' time from 23.6 minutes to 8.3, and a 2-minute test lets the plant ship all 11.0 orders.
- When both are stuck, add to both. With six pickers and three packing stations, the warehouse misses 1.1 orders.
- Add someone to supply parts rather than hide the shortage with more stock. With three kanban cards the line still stops for 36.6 minutes, but with two water spiders it stops for 0.0 minutes.
3. The articles
- How many pickers and packers?
- Is it reception or the docks that keeps trucks waiting?
- Where is the plant's bottleneck?
- How many water spiders?
4. How to ask an AI
In this warehouse (as drawn), for the afternoon's shipping, compare the orders that miss the 15:45 cutoff with four to seven pickers and two or three packing stations, 20 runs each, and tell me where the bottleneck is.
5. What this does not cover
- The orders and trucks that arrive, and the times to pick, pack, test and supply parts, are each article's assumptions. Replace them with your own site's records.
