Speak and done: What Voice to ERP requests look like
For twenty years now, process digitization has been a near-miss for employees working on the road. The mechanic, the driver, the inspector. What happens: the registration is added in the evening, or not at all. Voice-driven AI agents remove that friction: speak it in, the agent transcribes, links, creates the work order line, and schedules follow-up actions. The paradox: to become smarter with AI, you need better data — and you get that by lowering the barrier to record it.
Robert Keunen
4 min
There is one type of employee for whom every process digitization of the past twenty years has almost, but not quite, worked: the one who works out of the office. The technician, the service engineer, the driver, the inspector. Dirty hands, laptop in the van, no patience to fill out an entire form on a phone with a small keyboard.
What happens: the report is added in the evening, half-remembered. Or not at all. And that is a more expensive problem than it seems, because precisely the data that is most valuable — what happened on site, what was repaired, what the customer said — is what gets entered into the systems the worst.
Hold the button and speak.
New way of working: the technician opens the app on their phone, holds down a button and speaks:
“I replaced the pump at customer De Vries, serial number ending in 847, installed new pump, updated maintenance log, advised to replace the filter next time”
That's all. They release the button.
What the agent does next
In the seconds that follow:
- The speech is transcribed
- Customer De Vries is linked to the correct relation in your ERP
- The pump with serial number ending in 847 is searched for in the installed base
- A work order line is created, linked to this installation
- A maintenance action is recorded with date and technician
- The advice about the filter is converted into a follow-up appointment for the next cycle
The technician sees a confirmation on their screen. Done. No form filled out, no fields selected, no incorrect input on a small screen.
Applications outside the workplace
This principle — speak, agent processes — does not only work for technicians. Wherever people work on location or are on the road, the same friction exists:
• Account managers. After a customer meeting, dictate the summary from the car, and have it land directly as a contact moment in the CRM part of your ERP. No more hassle with loose notes that are only typed up a day later.
• Drivers. Confirm a delivery with a short comment about the location or recipient — important for the next route, but usually lost.
• Inspectors. Dictate findings with an attached photo, directly into the report. Usually, inspectors sit late at night typing out everything they saw during the day.
• Board of Directors. After a meeting, dictate the action points so that they are immediately created as tasks in projects — instead of being half lost in a notebook.
Why this is important
It is not about “faster”. It is about: registration without friction.
The lower the threshold to record something, the more complete your data becomes. And complete data is exactly what your AI agents need to deliver better work. An agent summarizing service history is only as good as the history that is in there. An agent using customer conversations to prioritize prospects only works if those conversations are actually recorded.
Less typing. More data. Better processes
The paradoxical rule: to get smarter with AI you need better data. To get better data, you need to lower the threshold for capturing data. And you do that not by making your forms prettier, but by scrapping them.
Speak and you're done. That is the simplicity the workplace has been waiting for for twenty years.