Your First AI Project in HR: Start With the Next Resignation
Most HR teams never pick a first AI project. Start with the next resignation: small, with an end date, easy to measure. A four-week plan.
HR teams are told to start using AI. Most never pick a first project. The usual suggestions are too big or too risky for a first try. There is a better place to start: the next person who resigns. A handover is small, it has a fixed end date and you can check whether it worked. Nobody's job depends on it.
Why first AI projects in HR stall
The advice is always the same. Find a use case, run a pilot, train the team. The trouble starts with the use case. The ideas that come up first are screening candidates, a chatbot for HR questions or predicting who will leave.
Each of these is a poor first project. Screening and attrition models make judgments about people. The EU AI Act lists AI used for recruitment and selection, and for decisions about promotion, termination or performance, as high-risk. That brings extra obligations before you have learned anything. A chatbot needs a complete, current set of HR policies to answer from. Most companies find out during the project that they do not have one.
So the pilot gets postponed. Or it starts and nobody can say after three months whether it worked.
Four tests for a first project
A good first AI project passes four tests.
It is small. One person, one team, a few hours a week.
It has an end date. You know in advance when it is over.
You can measure it. There is a clear answer to the question "did it help?"
HR already owns the process. Nobody else has to agree before you start.
Most of the popular ideas fail at least two of these.
Why a resignation passes all four
When someone resigns, a clock starts. The notice period is the end date. The work is small: one role and one person who knows it. The result is easy to check, because a successor arrives and either finds what they need or does not. And offboarding is already an HR process.
There is a second reason. What the person who leaves knows is mostly not written down. Some of it is tacit knowledge, the know-how people use without putting it into words. Some of it is the institutional knowledge of the role: why things are done the way they are. Getting that out takes good questions and a lot of writing up. Both are tasks where AI helps.
The project also makes no decision about anyone. It is about the work. That keeps it well away from the uses that make HR teams nervous.
How to run it in four weeks
Week one: pick the role. Choose one resignation in a role that is hard to fill or where one person holds most of the knowledge. Agree with the manager who runs the sessions. The first 72 hours after a resignation are the best moment to start.
Week one: prepare the questions. Give an AI assistant the job description and ask it to draft questions about the work, the people and the exceptions. Compare the result with our 42 knowledge transfer interview questions and pick the ones that fit. Use only an assistant your company already allows for work data.
Weeks two and three: hold the sessions. Book two or three sessions of about 45 minutes. Ask the leaver whether you may record them and say who will hear the recording and when it is deleted. Use the transcript as the raw material. The Handover Kit has a booking email you can send as it is.
Week three: let the AI draft the write-up. Give the assistant the transcripts and ask for a role document sorted by situation: what to do when the supplier is late, who approves what, which deadlines come back every month. Then let the person who leaves read it and correct it. The AI drafts. The people who know the work check.
Week four, and again at day 30: check with the successor. Ask the new person what they still could not find out. Their answers show what the handover missed.
This is a different conversation from the exit interview. The exit interview asks the wrong questions for this job, and it comes too late.
What to measure
Three numbers are enough.
Questions the successor still has after a month. Fewer is better.
Days until the successor handles a routine task alone. Compare with the last time someone joined that team.
Hours the manager spent. The sessions plus the checking. This is the number that shows whether AI saved time.
Write the numbers down before you start the next handover. The second one will be faster.
Rules to keep
Ask before you record. Say who will hear it and when it is deleted.
Use approved tools only. Transcripts contain names, clients and internal details. Keep them in the tools your company has cleared for that kind of data.
Follow your data protection rules. In the EU and the UK, recordings and transcripts are personal data under the GDPR and the UK GDPR. Agree a retention period.
Let a person sign off. The leaver checks the document before it is shared.
What to tell your team
Tell them what the project is for. It keeps what one person knows about a job, so the next person can do it. It does not judge anyone's work. The person who leaves helps write the result and checks it. Most people are glad to be asked what they know before they go.
Frequently asked questions
What is a good first AI project for HR?
A knowledge handover for one resignation. It is small, it ends with the notice period, it can be measured through the successor and HR already owns offboarding.
Do I need special software to start?
No. A general AI assistant that your company allows is enough for a first project. Specialised tools save time once you run handovers regularly.
Is this covered by the EU AI Act's high-risk rules?
The high-risk list for employment covers AI used to recruit, select, promote, dismiss or evaluate people. A handover project captures knowledge about the work and makes no decision about a person. Check with your legal or privacy team before you start, as with any new use of AI.
How long does it take?
About four weeks for the first handover, alongside normal work. The manager spends three sessions of 45 minutes plus time to check the write-up.
Want the next handover to run without the extra work? See how Rinto interviews the people who leave, asks the follow-up questions and writes the role document for the next person.
Sources
Regulation (EU) 2024/1689 (Artificial Intelligence Act), Annex III, point 4: employment, workers' management and access to self-employment.