A workday at Agrowteam, a Dutch builder of polytunnels and greenhouses. New greenhouse, a bill of materials from Autodesk Inventor with thousands of components, and someone spending hours figuring out what gets bought, what gets produced, and how it ends up in an Odoo BOM. Until recently, that was just the job.
Now AI does it in minutes. Same BOM, same outcome, a day less work per greenhouse. That is what AI in manufacturing can do, without needing to be revolutionary. One step that took hours, now goes in minutes.
This piece is about where else AI works in a manufacturing business, how we do that, and when to wait a bit.
What AI does in a manufacturing business
- Turn a CAD bill of materials into an Odoo BOM
- Flag the machine likely to fail
- Read supplier confirmations and delivery notes
- Give the planner suggestions
What AI does on the floor and in your costing
| What AI does | How it works | What you get |
|---|---|---|
| BOM skill | Reads a bill of materials from Inventor or another CAD package and turns it into an Odoo BOM, with the right make-or-buy choice per part. | At Agrowteam: from hours of work per greenhouse to minutes. |
| Predictive maintenance | Watches the maintenance log and flags the machine likely to fail. | Less unplanned downtime. |
| Document intake | Reads incoming supplier confirmations, delivery notes and invoices and sets up the booking. | Less re-typing, fewer errors. |
| Help the planner | Suggests moves when demand or a delivery date shifts. | Faster response, better utilisation. |
For none of the four do you have to produce new data. BOMs, work orders, stock moves and maintenance already live in Odoo.
How we approach this: one step at a time
At Agrowteam it was not an “AI strategy”. It was a concrete problem that kept coming back: a BOM from Inventor with thousands of components, turned over by hand. We built an AI skill that reads the BOM and prepares the Odoo BOM. The foreman reviews, corrects where needed, and pushes it into production.
What you see there is what we do more often: one job that comes back weekly or daily, automated cleverly enough that the person stays in charge. Not replacement, but relief.
In Updoo Shopfloor we showed this another way. In a 40-person manufacturing team paper work orders disappeared completely, and on-time completion went up 11%. AI sits naturally on top: suggest the next step, catch an anomaly. The data is already flowing through it.
At a large Dutch national printer the planning runs at a different scale entirely. More than sixty thousand orders a year, split across fifty workstations. Until recently that was largely manual, in Excel - with all the puzzle moments and errors that come with it. We are putting AI in there for an optimal daily plan on the orders already in Odoo, and Updoo makes sure that every employee at their workstation sees on a simple screen what they need to work on. AI plans, Updoo presents, people work.
Another example we are currently working on: a powder coater with three production lines who optimises his orders by hand every day. Colours need to be grouped and run in a fixed sequence, light to dark - otherwise the cleaning between batches eats more time than the coating itself. Going through orders one by one, day in day out, and a rush order halfway through throws the whole plan. AI can produce the optimal schedule for that day in one pass, reading the orders already in Odoo. Something slips in? AI takes the planner’s manual adjustment as a hint and runs a second pass, instead of someone losing another hour to it.
Even for service after the factory gate, the same idea works. In RogerDone we built AI scheduling for technicians from tasks generated in Odoo. What you sold, you know. Who can do what and where they are now, you know. AI puts a planning on top of that.
Manufacturing checklist: when does AI pay off?
| Start with AI when… | Wait a bit when… |
|---|---|
| …the process runs steadily and the data in Odoo is clean. | …your key data still sits loose in Excel or on paper. |
| …the task is narrow and high-frequency (one step, one type of document). | …it only happens once a quarter. |
| …you can measure the result in time, stock or errors. | …you cannot say what a good result is. |
| …you understand the job well yourself first. | …you cannot explain it well yet. |
AI on a messy process mainly speeds up the mistakes. Get the basics right first.
Why manufacturing custom work fits the budget now
The BOM skill for Agrowteam would have cost tens of thousands ten years ago, if anyone had built it at all. Too narrow, too specific, too little reuse. Today a prototype is a few days of work.
The work you pay for shifts. Less typing code, more thinking through what the thing has to do, and then testing it carefully. People call it vibe coding: AI does the typing, we do the craft. So small, focused solutions suddenly fit a mid-market budget. Not “a big AI platform”, but “that one step that always trips us up”.
For the wider story behind this: custom software is affordable for mid-market now.
Get your production data straight first, then AI
Walk your own factory floor for a day. Write down where the same look-up happens twice a week. That is your first AI job. Nothing more.
A few places to read on:
- Odoo for manufacturers - how Odoo lays the base that AI stands on.
- Can Odoo really run your production floor? - where Odoo stops and Updoo Shopfloor begins.
- Custom software is affordable for mid-market now - the full story behind the why.
- Run a wholesale operation alongside? Read AI in wholesale too - many patterns overlap.