Customer case study
How HR2day answers faster and better, on the most sensitive data there is
HR2day manages the salary, contract and absence data of the employees of more than two hundred organisations. That is exactly why they chose to adopt AI in a controlled way rather than wait or ban it. The result: their service desk now answers from their own documentation, with the source attached, while personal data is protected automatically along the way. And what started internally is now being rolled out to the specialists at their own customers.
The challenge
They wanted to use AI, and rolled it out in a controlled way
HR2day runs the personnel administration of other organisations. Marco Boerlage has been actively steering AI adoption for two years: monthly sessions with the whole team, ground rules for employees, high awareness. It went well. The point is that "well" based on human vigilance does not scale.
The worry is not abstract. At the municipality of Eindhoven, employees shared residents' personal data with ChatGPT. Marco's question: what happens when someone downloads a report full of address, salary, contract and health data and asks a public AI for a trend analysis? The next day someone else asks what senior management earns, and the AI answers. You cannot prove what was uploaded, cannot see who did it, and cannot get it back out.
Blocking was technically possible, but then you miss the improvement. For Marco that risk weighed heavier. And there was the manual work: a customer report he was enthusiastic about turned out to have taken two days, largely because all the data had to be made unrecognisable by hand. That ruled out repeating it.
“I see AI as an enabler for us to be better, to be faster, more tailored to our customers. That's why we do it. And at the same time we have to be careful, because we're dealing with personal data. That's the tension.”
The core
Why shielding everything does not work
Blanking out everything is very safe, and immediately useless. An employee calls in sick, you ask how to build a good absence file and you mention the role, the start date and the salary. No name given, but on a small team everyone knows exactly who it is about. And in a salary analysis, redacting the salary is the same as throwing away the question. So it is not about shielding as much as possible. It is about understanding the goal of the question: protecting what needs protecting, and letting through what is needed to answer it.
“The one time you have to protect the salary, in the other situation you don't want that at all. So it's very important for a tool to understand the context and the goal. Help me protect the things that need to be protected, and at the same time help me answer the question.”
The solution
The work gets faster. The protection takes care of itself.
The service desk answers customer questions from more than a thousand curated documents of their own: the question goes in exactly as the customer asked it, and what comes out can be checked and forwarded, even live during a phone call. Everything runs inside their own European cloud environment.
The source is always attached. The underlying documents are one click away. For a service desk that sends that answer to a customer, that is the reason to trust it.
The best AI model for every task. The right AI is chosen per question. No dependency on a single provider.
Protection before the AI looks. Sensitive data is protected before any AI sees it. Context determines what is needed, and what was protected returns to its place in the final result.
A reliable logbook, for customer and auditor. Who asked what, what happened to it and what came out is recorded. From "I cannot prove it" to "I can show you".
“I find that a billion times more powerful. And more fun, and simpler, because you don't have to think about it. It just goes right.”
What it delivered
From experiment to indispensable, in months
The report that took two days has become a repeatable process. The service desk helps customers faster, with the source attached. And the attitude flipped from being concerned about AI to being enthusiastic about it.
The next step
From internal to HR2day's own customers
Power users at HR2day's customers now get direct access to their own workspace with curated documentation, through the existing platform, without a second login. What started as the question of how to let employees work with AI safely has become a way to bring HR2day's knowledge straight to the people who work with it every day.
To be fair: trust in the answers is something you build, especially with employees who are new to it. That takes good source documentation and the habit of checking before anything goes out the door. HR2day and Aimable work on that together. “Quality over speed, a long-term view. That is what makes it work.”
Insight into what happens
With a public AI you cannot see what was done with your data. Here you can: requests and processing are logged. That also shows which questions live in the organisation and where people need more guidance.
Not locked to one AI provider
The right AI is chosen per question. A change at one provider does not stop the work.
Ready for the EU AI Act
Because use, processing and results are recorded, the platform helps HR2day comply with the AI Act.
“A great tool sells itself. You don't have to explain it, just use it and you'll see what it can do for you. And in a year's time we simply can't do without it anymore.”
Curious what this looks like for your team?
Book a demo and talk to Pim Verschoor directly. Pick a time in his calendar and he'll show you live how work that takes days shrinks to hours, safely, even with sensitive data.