Automating HR processes and onboarding: what works and what doesn't

Automating HR processes pays off most for work that runs the same way every time: getting a new employee set up, checking leave against balance and staffing, tracking deadlines during sick leave and answering recurring questions from your own policies. Conversations, judgements about people and anything involving medical data stay with people. Start with onboarding: that's where most of the separate tasks are, spread across most departments.

We've written these HR tasks out as AI employees for HR: per task what runs automatically, with which systems, and where a person decides. This article shows how to choose, what automating onboarding looks like step by step and which limits the law sets.

Which HR processes can you automate?

Processes with fixed steps, clear rules and lots of handover between people. The more departments involved, the more it pays off.

• Onboarding: from signed contract to a workspace that's ready. See the Onboarding Assistant.
• Offboarding: the mirror image, where the key step is revoking accounts and access on time.
• Leave: checking requests against balance and staffing, and only escalating the conflicts. See the Leave Request Handler.
• Sick leave administration: tracking the administrative steps and deadlines without touching medical data. See the Sickness Absence Administrator.
• HR questions: answers from your employee handbook or collective agreement, with the source included. See the HR Helpdesk.
• Recruitment: summarising and sorting CVs for a recruiter. Stricter rules apply there, see CV screening with AI.

How do you automate onboarding?

You define the sequence of tasks that follows a signed contract, let an AI employee assign and chase those tasks, and keep the welcome and the actual onboarding with people. The starting point is a single event in your HR system: a new employee with a start date.

1. The new hire as the starting point

As soon as a new employee is in your HR system with role, department, manager and start date, the sequence begins. In AFAS Profit, Nmbrs or Loket that's a fixed moment an integration can pick up. So nobody has to send an email to kick off the process.

2. Tasks per department, each with a deadline

HR, IT, facilities and the manager each get their tasks, counted back from the start date: ordering a laptop, creating an account, arranging an access pass, drawing up the onboarding schedule. The AI employee tracks what's done and chases what's stuck.

3. Accounts and access per role

Which systems and folders someone needs depends on their role. Define that once and the request can go automatically. How employees flow from AFAS to Microsoft 365 and Entra ID is covered in connecting AFAS to Microsoft 365.

4. Documents and the first day

The AI employee requests what's still missing, such as the signed contract or details for payroll, and flags a few days before the start what isn't ready yet. That leaves time to fix it, instead of on the first day.

Offboarding runs the same sequence in reverse. There, the last step matters most: whoever leaves loses access on their last working day. A forgotten account is a security risk.

What do you need to arrange before automating HR?

Three things: one source for employee data, written-down rules and an owner per process. Without those three, you automate today's lack of clarity.

• One source: your HR system leads for name, role, department and start date. Other systems take it from there.
• Rules on paper: which role gets which access, which documents are needed, when leave clashes with staffing.
• An owner: someone in HR who maintains the rules and handles the exceptions.
• Policies that are up to date: an HR helpdesk answers from your handbook or collective agreement. If that's outdated, so are the answers.

What shouldn't AI do in HR?

Make decisions about people on its own. The AI Act classes AI systems used for recruitment and selection, and for decisions on terms of employment, promotion, termination, allocating tasks based on behaviour or personal traits and evaluating performance, as high-risk (European Commission, AI Act Annex III, point 4, checked on 27 October 2026). Strict requirements apply there, including on human oversight.

Onboarding, leave based on balance and answering questions from your policies are different: the AI doesn't decide about someone's job or evaluation. Keep that line in the setup too. The AI employee prepares, a person decides. What else to arrange for the AI Act and GDPR is in the AI Act and GDPR: what do you need to arrange?.

With sick leave the line is even sharper. Medical data belongs with the occupational health doctor and service, not with HR, and so not in a system HR uses. What you can automate are the deadlines. In the Netherlands, according to UWV, the occupational doctor draws up a problem analysis when an employee has been ill for almost six weeks, and the action plan must be in place no later than the eighth week of absence (UWV, in Dutch, checked on 27 October 2026). Tracking deadlines like these is exactly the kind of work that can't be forgotten.

Does HR automation work with AFAS, Nmbrs or Loket?

Yes, through the package's API. Processing stays in your HR system itself, so data and balances are in one place. The AI employee reads what it needs and prepares what a person then confirms.

What's possible per package is covered under AFAS, Nmbrs and Loket. If new employees also need to be able to log tickets with the service desk straight away, see connecting AFAS to TOPdesk. We can connect any system, including an HR package not listed here.

Automating HR at staffing and recruitment agencies

Staffing and recruitment agencies deal with many more people a year, often on short contracts. Onboarding there doesn't happen a few times a year, but every week. Then automating contracts, documents and registrations pays off sooner.

What's involved there, from placement to invoice, is in what a staffing agency can automate and on the page about recruitment and staffing agencies.

Where do you start?

Count how many people join and leave each year, how many tasks belong to one onboarding and how many people they sit with. Then look at where it goes wrong now: a laptop that isn't there on day one, an account that survives someone leaving, a deadline that gets missed.

Pick one process with a clear owner and have the AI employee only suggest and remind at first. Want to know first which HR work costs the most and where automating really pays off? The AI Opportunity Audit maps that. After that, AI implementation helps with the foundation and the first process.