Onboarding New Staff Faster With AI: What Helps and What Stays Human

Key Takeaways

• AI helps onboarding most by letting new people answer their many small questions themselves, with a link to the source.
• An onboarding assistant is only as good as the knowledge that's written down.
• A go-to person, shadowing real work and feedback stay with people.
• Through the assistant, new staff only see what their role needs, and you analyse their questions anonymously.
• Every onboarding reveals gaps you can close before the next one.

AI speeds up onboarding mainly because new staff can answer their many small questions themselves: where a template is, how a process works, what an internal abbreviation means. An assistant that answers from your own documents and shows its source takes pressure off your experienced people. What AI doesn't replace is the go-to person, shadowing real work and feedback.

Where does onboarding time go?

Rarely into the planned training sessions, mostly into the countless small questions of the first few weeks. Each one is quick to answer, but each one interrupts somebody. New people also don't like asking a third time, so they search on their own, often for a long time and sometimes in the wrong place. The onboarding folder from three years ago doesn't help much either, because processes and filing have changed since. So how well onboarding goes depends on how much time your experienced people can find alongside their own work.

What can AI take over in onboarding?

Mainly the repetitive part of passing on knowledge: the explanations your experienced people would otherwise give from scratch to every new hire. Specifically:

• Answering questions: an assistant knows your handbook, processes and templates and answers with a link to the source, patiently, even the tenth time.
• Drafting an onboarding plan: a template per role becomes a plan for the first weeks, which the go-to person adjusts.
• Explaining terms: internal abbreviations, product names and the jargon of your industry.
• Practice: role-playing a customer call, a tenant complaint or a phone inquiry, with feedback based on your own guidelines.
• Summarising files: whoever takes over a running project or mandate gets an overview of the key documents and open points.
• Checking understanding: short questions about processes that show where there's still uncertainty.

What stays with people?

The go-to person, the culture and judgement. AI can explain what a process looks like, but not why you handle a particular customer differently. It doesn't build trust and doesn't notice when someone is overwhelmed. So keep planning a named go-to person, regular short check-ins and shadowing on real jobs. The assistant makes more room for those conversations, because the routine questions have already been answered. And new people learn a great deal in passing, for instance by hearing how a colleague handles a difficult phone call. No assistant can replace that.

What does an onboarding assistant need to be useful?

Above all, knowledge that's written down, because an assistant can only answer from documents that exist and are correct. How to collect and maintain that knowledge is covered in building an internal knowledge base. For onboarding you need at least:

• The key processes for each role, briefly described.
• Current templates and internal rules in one place.
• Access rights by role: through the assistant, new staff see what their job needs and nothing more.
• Someone who checks the answers and fills the gaps.

Think about the new person's own data as well. If the tool stores which questions someone asks or how they do in practice exercises, you're processing employee data. Under Article 328b of the Code of Obligations, an employer may only process such data where it concerns the person's suitability for the job or is needed to perform the employment contract. So analyse questions anonymously, to find gaps in your documentation, not to judge individuals. There's more in AI and the Swiss FADP.

What does an onboarding plan with AI look like?

It alternates between self-study with the assistant, shadowing real work and conversations with the go-to person. Here's a general pattern for the first month, which you adapt to each role:

1. Week one: settling in.Getting to know the systems, folders and contacts. The new person asks the assistant their first questions and learns to read the sources. There's a short daily check-in with the go-to person.
2. Weeks two and three: shadowing.Joining real jobs, then using the assistant to read up on what they've seen. Practising recurring conversations in role-play.
3. Week four: first tasks of their own.Handling standard tasks on their own, with the go-to person checking the results. Open questions are collected rather than asked one by one.
4. End of the first month: looking back.A conversation about what went well and what was missing, including the questions the assistant couldn't answer.

How do you introduce AI-supported onboarding?

In five steps, starting with what your most recent hires went through:

1. Collect the questions of your last new starters.Ask the people you hired most recently what they most often wanted to know in their first weeks and where they couldn't find anything.
2. Match or write the answers.Link every question to the document that answers it. If there isn't one, the responsible person writes a short how-to.
3. Set up and test the assistant.Test it with the questions you collected. Every answer shows its source, and when it isn't sure, the assistant refers people to their go-to person instead of guessing.
4. Adapt the plan to each role.Fit the pattern above to each role. On day one, show new people how to ask the assistant good questions; the basics are in prompting basics for office teams.
5. Improve after every onboarding.Questions without a good answer become new documents. That way each onboarding is a little easier than the last.

An example: a fiduciary firm hires a payroll specialist in January, right in the busiest season. Instead of taking every question to the team lead, the new person asks the assistant how joiners and leavers are recorded, where the template for salary statements is and which clients have special arrangements, each time with a link to the internal guide. The team lead sets aside a quarter of an hour a day for the questions the assistant couldn't answer, and then adds to the guides.

How can you tell onboarding is getting faster?

From a few simple observations that you compare between your last hire and the next one:

• When new people start handling standard tasks on their own.
• How often they interrupt experienced colleagues in the first weeks.
• How many questions the assistant couldn't answer, and how quickly those gaps were closed.
• What new starters themselves say about their onboarding after a month.

What it looks like when your whole team finds knowledge on its own is on the company knowledge page. If you'd like to build up AI use in your team deliberately, we support that with our team training, or you can get in touch straight away.