From the defective part to the work order in the ERP

Picture an ordinary afternoon at an auto parts plant.

The operator on line L2 sees a defective part come off the press. On the line tablet, he opens the occurrence form and photographs the part with its production order label. The AI reads the photo and fills in the form: production order 4471, line L2, assembly sector. He completes it by voice:

“Problem with the bearing, looks like the 6204.”

The AI agent replies right there on the tablet:

“Occurrence recorded: PO 4471 · line L2 · assembly sector. Symptom: 6204 bearing. Should I open a corrective maintenance work order in the ERP and notify Cassio, the shift mechanic?”

Operator: “Confirmed.”

“Corrective maintenance work order #8839 opened in the ERP. Cassio notified in Teams. I’ll let you know here as soon as he picks it up.”

Trail · record #45219.

Photo, report, work order opened inside the ERP the plant already uses, shift mechanic notified — and every step recorded in a tamper-evident trail.

Before that “opened in the ERP,” what happened is what still happens in nearly every plant in the region: the operator wrote it in a notebook or dropped a photo in the team’s phone group, told the supervisor at lunch, and the supervisor opened the work order manually in the afternoon — when he remembered. Between “I saw the defect” and “work order opened,” hours passed. Sometimes the entire production order ran through before the work order was ever created.

Why this isn’t a chatbot

The difference between what the agent does and what a chatbot does is one of kind — but it’s easy to confuse, because the conversation looks the same.

A chatbot reads text and responds with text. The cycle ends inside the conversation.

The F7 KORE agent receives the report — form, photo, voice — and acts inside the ERP. It opens a real work order, with a real work order number, assigned to a real mechanic, tied to a real production order. The cycle ends inside the actual operation, not inside a chat window.

Three things have to work for this to function:

• Record the fact at the source, the way it happens. A photo of the part read by the AI to fill in the form fields, a voice report in colloquial Portuguese, the production order number taken from the label instead of typed from memory. The data goes in once, from the person holding the part.
• Know the company. “PO 4471 is running on L2” is not data a chatbot has — it’s data that lives in the ERP, in the shift production queue, in the equipment history. F7 KORE is integrated with those systems and sees that context.
• Act with permission. Can the L2 operator open a corrective maintenance work order without supervisor approval? That depends on the company’s access rules. The agent executes under the permission of whoever confirmed — and goes no further.

That’s the difference between an AI that converses and auditable operational execution: the action happens in the system of record, and who authorized it is provable.

The channel is wherever the person already is

On the line, the operator uses the bench tablet, because the shop floor doesn’t always allow personal phones. The supervisor is in Teams, because IT standardized on Teams. The maintenance manager follows along by email or in F7 KORE itself.

In all of those places, the agent is the same and the action is the same. The only thing that changes is where the person talks to it.

So what matters isn’t how the report arrives. It’s:

• What the platform knows about your operation (ERP data, the equipment register, the history of occurrences and maintenance).
• Which rules the company has defined (standard procedure, approval policy, shift escalation).
• What the real integration is with the system where the action happens — not “an API the ERP exposes,” but “a work order opened with the right fields, in the right module, assigned to the right resource,” with every call monitored.

When those three are in place, the channel stops being the problem. The execution cycle is the same on the tablet, in Teams or by email.

Permission as design, not an option

Here is the difference between AI acting inside operations and an assistant that only suggests things.

The agent only executes within the permissions of the user who triggered it. It doesn’t inherit admin permission. It has no permission of its own. If the L2 operator cannot open a corrective maintenance work order without supervisor approval, the agent doesn’t open it either — it creates the approval request, notifies the supervisor and waits for their “OK” to come through their own channel.

And every confirmation is recorded — not as a debug log, but in a tamper-evident trail, where any change is detectable and which any auditor can verify without relying on F7:

• Who recorded the occurrence.
• What the AI read from the photo and what the operator corrected.
• What action was suggested.
• Who confirmed it, and with what permission.
• What action was actually executed, in which system, with which number.

That record is what separates AI applied to industrial operations from a demo experiment. Without it, any automated action becomes a liability on the first day of an audit.

What changes in operations

Three direct effects:

• The work order goes out while the defect is still on the line. The gap between “I saw the defect” and “work order opened” no longer depends on someone remembering in the afternoon. In corrective maintenance, that’s direct margin — a stopped machine costs by the minute.
• Informal coordination recedes. Notebook, lunch, a photo lost in a group chat, “I’ll tell you later” — all the word-of-mouth traffic becomes a record with an owner and a trail.
• The operator gains leverage. Instead of waiting for the supervisor, they trigger the system directly — within what their permission allows. When approval is needed, the request reaches the supervisor with the context ready: the photo, the production order, the symptom.

None of these gains require replacing the ERP — whoever the vendor is. They only require that the real action, inside the system the plant already uses, can be triggered by an agent with context, rules and the right permission.

---

F7 KORE is built by F7 KORE SOFTWARE LTDA, an InterSystems partner — including ERP integration in cases like Degraus (equipment rental) and Quinabra (regulated chemicals). Details on how it works on the FAQ page.

If you operate a mid-to-large industrial business, have an established ERP, and want to see a defective part turn into a work order in your own system — talk to us.