When I talk to manufacturing companies about AI, the first reaction is often the same: 'We want to start with it, but we don't know where.' And then almost always follows the suggestion to start with production. With the machines. With the operations.
That's understandable. Production is the core of the business. But it's rarely the best place to start with AI. And when I say that, people often look at me surprised.
Let me explain why.
The problem isn't in the machine — it's in the information around it
A CNC machine is running or it's not. That's fairly binary. The real chaos in a machine shop sits in everything that happens around that machine: who has the drawing? Is the material there yet? Which order has priority? What's the status of that rush job from last week?
Those questions are answered in most companies by a person who has everything in their head. The work planner who has known where everything stands for 15 years. The scheduler who knows from experience which customer wants to be called first. The purchaser who knows off the top of their head what the delivery time is from supplier X.
That's valuable — but it's also vulnerable. And it's exactly where AI makes the most difference.
Where AI adds direct value in practice
1. Work orders and documentation
How much time is lost in your company searching for the right drawing, the latest revision of a work order, or a customer's specifications? In most companies I see, that's at least half an hour per person daily. With a simple AI-driven search function on your own documents, that problem is largely solved.
2. Quotes and customer communication
Preparing quotes is a time-consuming manual process in many machine shops. Data is retyped from old quotes, specifications are processed manually, and each quote starts more or less from scratch. AI can take over a large part of the preparatory work here — without replacing your employee's expertise.
3. Purchasing and inventory monitoring
Systems that don't talk to each other, inventories that are manually maintained in Excel, purchase orders that are created too late because someone forgot. These are processes where automation with AI delivers direct, measurable gains — already in the first weeks.
4. Planning support
Not replacing the planner — that human expertise is irreplaceable. But supporting them: automatically signaling when an order threatens to run late, providing insight into capacity bottlenecks, or simply offering overview of the status of ongoing orders without someone having to search through three systems.
Why not start with the machine itself?
Connecting machines to AI systems is technically more complex, more expensive and riskier than people think. Sensors, PLC connections, real-time data streams — that requires a serious investment and a solid foundation. If that foundation isn't there, the project ends halfway through.
Moreover: if the information flows around the machine are already chaos, more data from the machine won't solve that. You'll just get faster insight into a chaotic process.
First understand the processes. Then automate. Then expand to the machine. In that order it works.
The first step in practice
Start with a concrete process that currently runs manually, is time-consuming and has little risk if it goes slightly differently. Not the entire work preparation at once — but for example: automatically looking up and summarizing relevant information for an incoming customer inquiry.
That sounds small. But if that works, you have something in your hands: a working example within your own company, on your own data, managed by yourself. That's the foundation on which you can build.
And that foundation — your own AI environment that runs on your own infrastructure — is exactly where we start.
Would you like to brainstorm about which process in your company is most suitable as a starting point? Schedule a no-obligation introduction.