How employees without technical knowledge use AI in your ERP
The biggest fear in AI projects in mid-sized companies isn't "will the technology work? "That's a fear for the IT vendor. The fear in the boardroom is different: "we don't have the people in-house who can build this."
Justified, if you think in terms of developers, API specifications and Python scripts. But that's not how it works anymore. The people who get the most value out of AI in your ERP aren't the programmers. They're the people who run your processes daily.
What it looks like in practice
Take a projectleader at an installation company. She has five contracts in front of her: all service subscriptions, all with slightly different conditions.
The old way: read each one, note the differences, mail legal, wait.
The new way: she selects the contracts in your ERP and types in the chat: "What deviations do you see from our standard contract, and what risks dothey contain?" Thirty seconds later, there's an overview with deviations, risk points and a draft mail to legal ready. No training. No script. Just typing.
What's at work here
Three steps,nothing more:
1. Select documents, requests or relations in your ERP
2. Choose an assistant agent from the bar
3. Give a command in the chat — in plain English
The agent knows which data you selected, knows your ERP system, and can combine the relevant information and prepare actions. What used to be an IT project —"we need to build a report that compares contracts" — is now a question the project leader asks herself.
More use cases across theorganisation
This pattern repeats on every department.
- Finance: a controller uses it to testinvoices from a batch against procurement conditions — deviations surfaceautomatically.
- Service: a planner summarises and categorises a stack ofcomplaints to route them correctly.
- Sales: an account manager searches the quote history of a specific customer to compose a targeted mail.
- HR: employee matches incoming CVs to avacancy profile and gets an initial screening.
In each of these cases, it's not the IT department steering the agent. The employee who knows the process does it themselves.
Why this is the fundamental shift
For a long time,automation was a sequence of investments: first an ERP, then a CRM, then a BI tool, then an integration platform. Each step required an IT trajectory,specialists and lead time.
With AI agents, that dynamic changes. The people who know the process — not the people who know code — steer the agent. Domain knowledge wins, not code. Anyone who can type a mail can operate an agent.
That doesn't mean IT disappears. On the contrary: someone has to configure the agents, manage rights, secure integrations. But daily use shifts to the shopfloor. And that's where the gain sits, because that's where the repetition lives.
What you can do tomorrow
No six-month pilot. No steering committee. Give three people from one team two weeks of access and simply ask them: where did it help you?
What works spreads itself through the organisation. What doesn't work, surfaces quickly enough too.
That's the most pleasant way to introduce AI in a mid-sized environment: not as a big transformation programme, but as a tool people discover themselves.
