Easing a new employee's onboarding with AI
Bringing someone new into a small company always goes the same way: in the first week the new person never stops asking, and the person who knows most never stops answering. "How do I issue a credit note?" "Where are this year's rates?" "Who do I tell if we are out of stock?" Every question is reasonable. The problem is that they are always the same ones, and that whoever answers is usually whoever has the least time.
In large companies HR solves this with manuals and induction courses. In a small company, the "induction manual" is usually a very patient person. And when that person is on holiday, or simply busy, the new employee gets stuck or improvises. Improvising on your second day is expensive.
The manual that answers when you ask it
The good news is that the same technology behind an internal knowledge base works for onboarding. The idea is simple: gather the company's procedures, documents and habits — how a quote is done, where the templates live, what the warehouse hours are — and put an assistant in front that answers in plain language.
The new employee, instead of interrupting a colleague for the fifth time that morning, types the question as they would say it: "how do I log a return?". The assistant answers with the company's own procedure — not a generic answer from the internet — and says where the original document is if they want the detail. Basic questions stop going through the team; questions of judgement, the ones that matter, keep going to people.
Not everyone needs to know the same things
A common mistake is giving everything to everyone. It makes no sense: a shop assistant needs the till, exchanges and returns; someone in administration needs to know how suppliers are invoiced and where the contracts are. A good onboarding system is set up by role: each profile sees the procedures and documents that concern it, in the order that makes sense. In the first days the assistant can even act as an itinerary: "today you learn how we handle orders; when you have it, ask me anything."
That also protects the company: the new person does not have to wander through folders containing everything, and the company knows exactly what each person has access to.
An example: the firm that hires someone in July
Imagine an accountancy firm taking someone on mid-season, with the team overwhelmed. Before, the manager spent half of each of the first five days training the new person: where every form is, how every filing is submitted, which clients have quirks. About twenty hours, at the worst time of year.
With an onboarding assistant, the firm documented its procedures once (most were already in PDFs and emails) and now the new person asks it directly. The manager went from twenty hours of explaining to three or four of real supervision: reviewing work, handling special cases, getting to know the person. And when someone else joins in a few months, the work is already done. The same holds for a workshop with rotating apprentices or a shop hiring for the Christmas season: the investment is made once and pays off at every hire.
The small print: the new employee's data (and everyone else's)
Here is a point that cannot be skipped. Sensitive data moves through an onboarding process: the contract and personal details of the person joining, and internal documents that may contain payroll, terms or client information. None of that can end up in just any free tool.
Three minimum conditions before putting any system in place: access controlled by role (the new shop assistant must not be able to ask about anyone's payroll), data processed and stored inside the European Union, and a contractual guarantee from the vendor that your documents are not used to train third-party models. If a tool cannot answer those three questions clearly, it is not for your company.
Start small
You do not have to digitise the whole company to begin. It is enough to write down the ten questions every new employee asks in the first week and document the answers. Only then does it make sense to put an assistant in front of them — and by then you will have discovered that half the value was in writing the answers down at all.
This article is general information, not legal advice for a specific case.
About this article
iaetica.cat is a space for writing in the open: what I am learning about artificial intelligence, its ethics and its law.
— Laura Graells