One message from the shop floor instead of ten phone calls. A concept for a chat that checks the data first

I know this scenario from a project I am developing right now for a manufacturing company. We are working on a concept meant to replace that chain of phone calls with a single message. I am describing it deliberately at the idea stage: this is the start of a series in which I will show how such a system takes shape step by step, with the decisions, the changes of direction and the lessons from going live.

A company chat with channels, with one difference

The starting point looks familiar: a chat with channels, much like Slack. Every channel has a host, meaning a person or a team responsible for one area: purchasing, process engineering, maintenance, stores.

The difference sits in what happens between sending a message and its arrival in a channel. The employee does not need to know who should handle his problem. He simply writes: “material missing for product XYZ”. Before that message goes anywhere, the system underneath, built on AI (artificial intelligence), classifies the topic and checks the data in the production system on its own.

What the system checks before it asks people

In the missing material example, the sequence of checks looks like this:

• is there a customer order for this product and what is its status,
• has the routing for this order been prepared and released to production,
• does the works order record the availability of tools and materials,
• if the material is missing: is there already a requisition in the system,
• if a requisition exists: has the material been ordered from the supplier,
• if it has been ordered: what delivery date is expected.

Now the branch that matters most. If the collected data settles the matter, the person asking gets an answer automatically, within seconds: “material ordered, confirmed delivery date: Thursday”. Nobody is pulled away from their work, and the call to purchasing never happened.

If data is missing, the system creates a message and routes it to the right channel or channels, where the people who can move the matter forward sit. And here the concept goes a step further than an ordinary chat: alongside the message, tasks are created for specific people, with priorities. If the product is urgent because the customer has a confirmed short shipping date, the task jumps to the front of the queue, ahead of other matters.

From checking to acting

Checking the data is only half of the concept. The system is also meant to carry out the next steps of the process itself, always with a human on the decisions that have consequences.

Back to the example. The material is needed, and the system holds no requisition yet, meaning the document from which the purchase order to the supplier is later created. In that situation the system raises the requisition itself and sends the responsible person a message in their channel: “a requisition for material X has been raised”. Two buttons sit next to the message: approve or reject.

Once approved, the system generates the purchase order for the right supplier and sends it for approval to the orders channel. That may already be a different channel and a different person from those for requisitions. Again a conversation in the chat, again buttons. When the host of the orders channel approves the document, the system sends the order to the supplier.

The system's work does not end there. It is also meant to follow the emails the supplier sends back confirming the order, and to find the expected delivery date inside them. It writes the date it found into the system and only then returns to the person who asked in the first place: “the material will be available on Thursday”. Along the way every step has its task assigned to the right person, and once a step is done, including automatically, the task is marked as complete.

The division of roles is simple: a human decides with one click, and the system prepares the documents, keeps the sequence in order and carries information between people, emails and production data.

The system closes the loop itself

In everyday practice the last step is the one most often lost: someone dealt with the matter but forgot to pass the word on. The material is ordered, and the operator at the machine is still waiting for an answer.

That is why in this concept the system watches the closing itself. If the task read “order the material” and a purchase order for that material appears in the system, the system spots it: it marks the task as complete and replies to the person who asked at the start, with the delivery date if that is already known. Nobody has to remember to report back, because the source of truth is the data, and the system reads it itself.

The end result: one message sent by an employee sets the whole flow in motion, meaning the data check, the requisition, the purchase order to the supplier after two approval clicks, the delivery date read from an email, the answer to the person asking, plus tasks that create and tick themselves off.

Why bother

In the EIB Investment Survey 2025, the availability of skilled staff is the second most cited barrier to investment among EU firms, named by 79% of them, and manufacturing is among the sectors where it bites hardest. The Eurochambres Economic Survey 2026, which gathers responses from tens of thousands of businesses across 28 European countries, puts a shortage of skilled workforce among the top three challenges for the year, next to labour costs and regulatory burden.

When people are hard to find, the time of those already on the payroll matters more. And a large part of that time in companies of 20-250 employees disappears exactly into the flow of information: who knows where the material is, who was supposed to prepare the routing, who should call the supplier.

A chat that checks the data first turns the proportions around: people handle the exceptions, and routine questions answer themselves.

What comes next

This is a purely conceptual stage. We have not decided yet whether to build on an existing chat tool or to build a dedicated one. The questions about the limits of automation are open too: which steps the system may carry out on its own, and which should always wait for a human to press “approve”. Those decisions will be the subject of the next articles in this series, along with what works and what has to go in the bin.

If you run a manufacturing company and recognise the same chain of phone calls, write to me. I would gladly compare how this problem looks in different plants, and take that into account before the concept becomes a system.

You will find more articles on production management and digitalisation without chaos in the Knowledge section.

Sources: EIB Investment Survey 2025, European Union overview; Eurochambres Economic Survey 2026.