"We're going to add AI to our ERP." It sounds good in the boardroom. But whether you lead a 200-person organisation or you're responsible for daily operations, you want to know what concretely changes on Monday morning — when the first mail comes in and the first work order needs processing.
Honest answer: a chatbot on top of your ERP changes little. Your people already knew how to search. It gets interesting only when AI can not only talk about your ERP, but also act within it.
From talking to doing
That difference lies in something called tool calling. You give an AI agent access to the integrations you've had for years — to Exact, AFAS, Business Central, your CRM, your document storage — and let it use them autonomously to complete a task. No human intermediary retyping a request, but an agent that chooses itself which action is needed.
An example. A mail comes in from a customer with a delivery note attached and a request to adjust a quote. Without agent: someone reads, looks up the customer, opens the project, reads the delivery note, adjusts lines, sends aconfirmation. Fifteen minutes, at best. With an agent: the mail is read, the delivery note extracted, the customer matched, the change prepared as a draft, a proposal mail ready for review. Two minutes.
That last step matters. The agent does the groundwork, a human signs off. That's where it works in mid-sized and industrial companies: not full automation where no one has grip any more, but smart preparation where the human keeps deciding.
Why this is possible only now
We've been building ERP integrations for years. That part isn't new. What is new: a language model can now decide itself which integration to call to answer a question or execute a task. Previously every path had to be programmed in advance — if this, then that. Now you can tell an agent "solve this customer query," andit chooses whether to retrieve project data, analyse a document or create a new relation.
This changes the strategic question.
For a director: not "which new system should we buy", but "how much additional return sits in the systems I already have, if an agent can use them autonomously?"
For the IT or operations manager: not "which integration should we build" — you already have it — but "which processes do we let an agent handle, and where do we keep humans in the loop?"
Where you can start tomorrow
Pick one process that costs your people a lot of time and requires a lot of switching between mail, ERP and a third system.
Purchase confirmations, Service notifications, Quote requests.
Choose one, not three. A good agent isn't built for an entire Department at once. You build it for one repetitive process and expand once it proves itself.
That's also the honest story behind AI in your ERP. No revolution, no self-steering back office.
But: years of integration work that can finally be used autonomously, process by process. And that's where the gain sits —measurable, controllable, and for the people who work with it daily, finally noticeable.
