A.I. that gives your HR team breathing room.

This is the question we get most often. Something new arrives every week, and that makes choosing harder, not easier.

We don't start with a tool, we start with your process. In an A.I. Opportunity Mapping we put your HR processes on the table together and look at where the time really goes. Only then do we pick a tool. Sometimes that's a tool you already have.

What we deliberately don't do: talk you into something because it's new.

Yes. That's the case at almost every HR team, and it's no reason to wait.

In the mapping we work with sticky notes and conversation, not with documents you have to supply first. At an HR department that started this way, half a day with eight people from HR, IT and digitalisation produced a ten-step process map. From the intake of a vacancy to onboarding. That map didn't exist anywhere before.

This is the first question HR directors ask us. The short answer: A.I. may help you sort and summarise, but the decision about a candidate stays with a human.

Three things to know:

• The GDPR does not allow you to base a decision with significant consequences solely on automated processing. A rejection falls under that.
• Collective labour agreement no. 38 requires you to treat all candidates equally and to respect their private life. An A.I. that sorts on the wrong characteristics gets you into trouble there quickly.
• The European AI Act places A.I. for recruitment, selection and evaluation in the high-risk category.

That's why we'd rather build A.I. that lowers the threshold than A.I. that filters. At a recruitment company we built a CV bot that helps candidates put together a good CV themselves, in their own language and on their phone. Not A.I. that rejects people, but A.I. that makes applying easier.

No, there's no need to wait. But do know where you stand.

The state of play in plain terms:

• What is already banned. Since February 2025 you may not use A.I. in the workplace that tries to read the emotions of employees or candidates. That includes a job interview.
• What is already required. Since 2 August 2026 you have to tell people they're talking to A.I., and A.I.-generated content has to be recognisable as such. For generative A.I. that obligation moves to 2 December 2026.
• What comes later. The heavy high-risk rules, including A.I. in recruitment, selection, evaluation and promotion, have been postponed to 2 December 2027 via the Digital Omnibus.
• What stays. The obligation to make your people A.I.-literate remains. That's exactly what training is for.

Our advice: don't use that time to wait, use it to do your homework. Know which tools are being used in your organisation, record who decides, and make sure you can explain what every A.I. step in your selection process does.

Often yes, and that gets forgotten.

Collective labour agreement no. 39 requires companies with at least 50 employees on average to inform the works council in writing when a new technology has significant collective consequences for employment, work organisation or working conditions. That has to happen no later than three months before it is introduced. Consultation follows. If there is no works council, it goes to the union delegation or the health and safety committee.

In practice: allow for those three months in your planning, and take them into your business case. We help you draft the note, but the decision and the consultation stay with you.

Yes. Copilot sits in Word, Excel, Outlook and Teams, and that's exactly where an HR team's work sits. Interview reports, work regulations, emails to candidates, presentations for management.

We will tell you honestly where Copilot is weaker than other tools, so you know when it makes sense to revisit your policy and when it doesn't. We're not tied to a single vendor.

A lot, because there's a great deal of template work there. Work regulations, contracts, addenda and policies are largely standard, with a small part that differs per client or per employee.

What we typically build there: have the standard part drafted automatically, leave the variable part as an open field, and build in a check exactly where the risk sits. You review what deviates, not everything.

Important: we are not employment-law advisers. We build the process; you or your payroll agency stay responsible for the content and for keeping up with the legislation.

It may, with the right arrangements. And it's one of the quickest wins in HR, because the follow-up work after a conversation usually costs more time than the conversation itself.

What you need to arrange: you say it up front, you ask for consent, you record how long the recording and the report are kept, and you make sure the candidate or employee knows they can request the report. We work with Leexi, where processing happens inside Europe.

What you get: a report in your own structure, straight into the right place in the file, with the agreements and next steps pulled out. For appraisals you also have a line over time for the first time. Where did we stand last year and what has happened since.

No, and that order is often exactly why everything stalls.

At an HR service provider hesitating between buying an existing ATS and having something custom built, we deliberately separated those two questions. We mapped the process, set out what a system really has to do, and compared four existing systems with a custom solution. Alongside that we surfaced eight improvements that could start immediately and that keep working in every scenario, including after a system switch.

That way you gain time while the big decision ripens, instead of waiting for it.

The applications we build or see working most often:

• Drafting job ads from an intake, in the company's tone of voice, ready for several channels.
• Helping candidates put together a complete CV themselves, so you get fewer half-finished applications.
• Recording and summarising conversations, with the agreements landing straight in the file.
• Sending an automatic message when a candidate moves to the next stage, so nobody is left waiting without news.
• Making every recruitment channel measurable, so you see where your good candidates really come from.
• An internal HR chatbot for the questions that keep coming back. Leave, time credit, policies, expenses.
• Preparing employment-law documents and addenda from templates.
• Onboarding: sending out the task list, the emails and the access rights automatically.
• A coach bot that supports managers between two coaching sessions, with escalation to a real coach.
• Sorting and labelling mail, so the HR mailbox becomes workable again.

This is the hardest question on this page, and it usually lands with HR because nobody else asks it.

Our honest answer, in two parts. First: the fear isn't irrational, so don't pretend it is. What does help is clarity. Say what you will and won't automate, and say what happens with the time that frees up. Freed-up time doesn't automatically mean fewer people, but somebody does have to say that out loud.

Second: what takes away the most fear is letting people use it themselves. In every programme we work with the participants' own cases, not with examples from a manual. Someone who sees for themselves that the tool takes over their dullest task and not their expertise talks about it differently.

We see that often. An organisation rolls out ChatGPT or Copilot, and a few months later half the team has stopped using it.

Usually that isn't the tool's fault. It's because nobody owned it and because too much arrived at once. We also almost always see two speeds in a team: a group running ahead and a group dropping out.

What does work: one internal owner per department, one process at a time, and a recurring moment where we look together at what's sticking. At clients where we skipped this, the team fell back on the old way of working after about three months. That's a lesson we learned ourselves.

It happens, and it isn't going away. CVs and cover letters look better than they used to and say less about the person.

Our view: stop using the written part as a selection filter and shift the weight to what A.I. can't fake. A good conversation, a short practical test, a conversation about real work somebody has done. A.I. can help you prepare and write up those conversations faster, so you can do more of them.

We advise against installing an A.I. detector. They're unreliable, and you risk turning away good candidates on the basis of a suspicion.

No. Anyone who can work with an HR package can work with this.

What we do instead of giving a generic course is sit down next to you. In the training we work with your vacancies, your documents and your files. That sticks better than a course.

What does take time is the beginning: sitting down together once to get your process and your settings right.

No. We're an A.I. partner, not an HR service provider. We don't give employment-law advice and we don't do recruitment.

What we do is make your service faster and better, and help you build new services on top of it. Several of our clients are HR service providers, payroll agencies and recruitment companies themselves. If part of your client base is starting to ask what you're doing with A.I., that's exactly the conversation we like to have.

A data processing agreement is a contract setting out what we may and may not do with your data. Who can reach it, where it sits, how long we keep it, and what happens when you stop. We sign that at the start of every engagement, not afterwards.

In plain terms: you stay the owner of your data. We're only the party working with it on your instruction. Your data is never used to train A.I. models, and processing stays inside Europe.

In HR this weighs even heavier, because you work with application files, salary data and appraisals. That's why we agree up front, per process, which data ends up where, and why we don't use public chat tools for personal files.

We prefer to start with a session where we map the opportunities together, then a business case where we cost out one idea. That way you know up front what it returns and in which order to begin.

A hands-on training for your team is the cheapest way in and saves time in the very first week. For a built flow you'll see a working pilot within four to six weeks and you're usually fully live after eight to twelve weeks.

We only take on projects we believe will return more than they cost in the first year.

We work in phases, with a decision point after each one. You start small, with one site or one process, for instance. If it works, we continue. If it doesn't, it stops there and you haven't paid for the whole programme.

Honest answer: there is one. A.I. doesn't free up time out of nowhere. The gain sits in the preparatory work, and that gain only becomes real money once you also decide what to do with the time freed up.

And in HR there's a second one: connections to HR software are sometimes tougher than a demo suggests. Many HR and ATS packages give you only limited access to your own data. That's why we measure up front how much time a process really costs today and check the connections before we promise anything. If the sums don't add up, we'd rather say so now than after three months.