On the first day the laptop is missing, the access is missing, or both.
A new colleague arrives on Monday morning. The desk is there, the welcome is warm, and the laptop is still stuck in an IT queue. Or it is there, but the ERP access was never requested. The first hour goes into apologies.
The reason is rarely sloppiness. The onboarding checklist lives in an Excel file and hangs on one HR person. She knows the plant badge takes ten working days and that controlling needs a BI license. When she is on vacation, that knowledge is on vacation. IT, the department, and HR coordinate by shouting down the hallway and by forwarded email threads. Nobody has an overview of what is still open.
After day one it continues. New employees ask the same questions for weeks: how do I book a business trip, where is the vacation request, who approves my expenses. They ask whoever sits closest, usually the new manager. That makes it take weeks longer until someone is really productive.
We start with the checklist that already exists.
The starting point is not an HR process model but the existing Excel file and the person who maintains it. That file is the most precise specification in the building. And the HR team’s conditions became the requirements: the conversations stay human, HR always decides the exceptions, and HR maintains the checklists itself. A common objection at the start is that onboarding is relationship work and a bot on day one feels cold. The objection is right, which is why the AI employee only takes over the coordination, never the welcome.
- Read the real checklist: Go through the existing Excel file line by line with the HR person who owns it, including the unwritten items: which access takes ten working days, which safety briefing is mandatory before entering the workshop.
- Start with one role family: The role with the highest volume, not the most complicated one. Its checklist becomes the first template, with a lead time on every provisioning item.
- Prototype within days: A status board per new hire: checklist scheduled against the start date, open tickets, missing documents, flagged exceptions. HR reviews it on real upcoming hires, not on demo data.
- Go-live only after the team’s okay: Answers come only from sources HR has released, always with the source shown. The system went live once the HR team said, after three real hires: this really saves us time.
80 percent is done by the AI employee. The human decides at two points.
The trigger is the signed contract in the HR system. From role, location, and start date the AI employee starts the matching checklist. A controlling hire gets different items than a field service technician. Every item is scheduled backwards from the first working day along its lead time. So the plant badge ticket does not go out three days before the start, but three weeks before.
From there three tracks run in parallel. The AI employee raises the tickets for laptop, licenses, and access in your IT ticketing. It asks the new colleague for the missing documents and follows up on schedule until the file is complete. And from day one it answers questions from the released company knowledge, always with the source document attached. The human stays in the loop at two points: HR decides every exception that cannot be derived from a rule, and HR welcomes the new colleague on the first day. The AI employee does not replace the welcome, it makes sure everything is ready by then.
New employee in controlling, starting in 3 weeks, headquarters location. Contract signed by both sides in the HR system.
One field is flagged for review: access to group reporting is not part of the controlling role profile. The AI employee therefore does not request it. HR and the department decide.
That was before, this is today.
This is how onboarding ran before, and this is how it runs today with the AI employee. Conservatively calculated, that saves about 80 percent of the HR coordination effort per new hire.
| Before | Today | |
|---|---|---|
| Effort per new hire | 4 to 6 hours of HR coordination | Under one hour for exceptions |
| Checklist | Excel file owned by one HR person | Role profile starts automatically, HR maintains it |
| First working day | Laptop or access missing | All ordered with lead time, ready before the start |
| Documents | HR remembers to follow up, or nobody does | Requested and reminded until the file is complete |
| First-week questions | Weeks of asking whoever sits closest | Answered from company knowledge, with the source |
* Baseline from process conversations with HR teams in the German Mittelstand. Savings conservatively calculated: before, 4 to 6 hours of coordination per new hire spread over several weeks for copying the checklist, raising tickets, following up, scheduling meetings, and answering questions. Today under one hour remains for exception decisions and approvals, about 80 percent less. The vacancy and ramp-up cost figure is a common industry anchor, not a measurement from a single project.
How do you build an AI employee like this, technically?
The knowledge from this scenario to take away, whether you build with us or on your own:
Integration: the HR system and IT ticketing stay the source of truth
The HR system delivers role, location, and start date, the IT ticketing receives the requests for laptop, licenses, and access, the document management holds the personnel file. Everything connects through the existing interfaces, with Microsoft 365 through the Microsoft Graph API for mail and calendar. There is no new platform and no second personnel file.
Checklists are a rule set, not an AI decision
There is one template per role family, plus items the location adds, for example the plant badge at a production site. Every item carries a lead time, and scheduling runs backwards from the first working day. This is deliberately a fixed rule and not a model, because nothing here may be guessed.
Answering questions: GPT and Claude, but only on released knowledge
For first-week questions, GPT from OpenAI or Claude from Anthropic fit well. They work exclusively on the handbook, policies, and forms HR has released, and every answer names the source document. If the answer is not in the released sources, the question goes to a human. A confidently wrong answer to a new employee is worse than no answer, because they cannot judge it yet.
Reminders: the logic that makes the difference
Missing documents are not requested once and then forgotten. The AI employee knows the due date and escalation level per document, reminds the new colleague on schedule, and reports to HR when a deadline slips or a ticket was not confirmed in time.
UX: the interface decides adoption
A status board shows per new hire what is done, open, or blocked. Reminders and emails open as an editable draft, nothing goes out unseen. And the question chat shows the source under every answer. Exactly these three things turned skepticism into approval.
What does it cost in comparison?
Superkind charges per use case. The price grows with the number of new hires, not with headcount. Here is the honest comparison:
| Excel checklist | HR software workflow | Superkind AI employee | |
|---|---|---|---|
| Cost | No license, but 4 to 6 hours of HR time per hire | 5 to 15 € per employee per month | Price per use case, a fraction of a full-time position |
| What is included | A list, everything else is done by hand | Workflow and task assignment, but no document chasing and no answers | Checklist, tickets, reminders, scheduling, and questions from company knowledge |
| Scales with | More HR hours | Total headcount, including existing staff | Number of new hires, without new positions |
| Exceptions | Sorted out down the hallway | At best an escalation field | Flagged and handed to HR with a reason |
| Rollout | Immediate, but re-explained every year | Configuration and rollout in the HR system | 2 to 3 weeks to the first productive version |
HR software delivers the workflow, and it does that well. But it does not chase a missing certificate and it does not answer questions from new employees. The honest comparison is the full cost of a rough start: the HR hours, the weeks until productivity, and the risk that someone quietly doubts the decision in week three.
What we learned from this scenario.
The reflex to protect onboarding from automation is a good reflex. But the part HR teams want to protect is not the part that gets automated. Nobody ever felt welcomed by an access request. What really damages the first week is the opposite of warmth: a manager spending Monday morning on the phone with IT instead of with the person they hired.
The second thing is surprising every time: how much undocumented knowledge sits in a single onboarding checklist. That the plant badge takes ten working days, that one department always answers late, that a specific safety briefing is mandatory. None of it is written down anywhere, all of it matters. Building the AI employee forces that knowledge into the open, and several HR teams said that alone was already worth it.
What it is not suited for: If you hire a handful of people per year, onboarding is not your bottleneck. Without defined role profiles the AI employee only automates ambiguity. And without a ticket system in IT it has no place to raise the request at all.
