Flitz.ai
← All articles
5 min read

What Happens to Your Know-How When Someone Hands in Their Notice?

A new hire, a departing employee, and a pile of undocumented processes — a common Swiss SME scenario shows why institutional memory needs to live somewhere other than people's heads.

What Happens to Your Know-How When Someone Hands in Their Notice?

Picture a small Swiss engineering firm with a dozen employees. It is late in the quarter, the bookkeeper who has handled vendor relationships and quarter-end close for three years has just handed in her notice, and a new hire starts in two weeks. This is a realistic, illustrative scenario — not a real company — but it is the kind of moment almost every Swiss SME eventually faces, and it exposes a problem most businesses don't think about until it's too late: where does the know-how actually live?

In this scenario, a lot of it lives in the bookkeeper's head. Which vendor needs a reminder email before the 20th of the month. Why a particular customer gets a different payment term. How the quarter-end MWST reconciliation checklist actually works, beyond what's written in the accounting software's help pages. None of this is a secret — it's just never been written down, because writing it down has always felt like a chore nobody has time for.

The 42% problem

This isn't a one-off. IDC research suggests that 42% of process knowledge inside SMEs exists only in employees' heads. It never makes it into a document, a wiki, or a handover note. When that employee leaves — for a new job, parental leave, retirement, or simply a sick day during a critical week — the knowledge leaves with them. The business doesn't just lose a person; it loses a working process.

Most companies know this is a risk. Many have even tried to fix it with a Notion workspace or a Confluence wiki. The intention is good. The outcome is usually the same: a handful of enthusiastic entries in month one, then silence, because writing a polished wiki article takes real effort and nobody's job description includes "maintain documentation."

How it plays out with a working knowledge bank

Back to the scenario. Instead of letting three years of institutional memory walk out the door, the outgoing bookkeeper spends twenty minutes before her last week doing something that doesn't feel like a chore: she opens the intranet knowledge bank and, for each process, pastes a short description — "how we handle quarter-end MWST reconciliation" or "our vendor payment reminder routine" — or even a URL to a reference doc she's already linked to internally.

The AI drafts a polished entry from that — title, summary, full body, tags, and category — in seconds. She reviews it, corrects a detail or two, and posts it. No blank-page problem, no formatting effort, no excuse to put it off until "next week."

Two weeks later, the new hire starts. Instead of shadowing a colleague for the first month and asking the same questions in Slack three times, she asks the knowledge base directly, in plain language: "How do we handle a customer who disputes an invoice?" or "What's our process for reconciling the quarter-end books?" The system answers with citations pointing back to the exact entries it drew from — so she's not just getting an AI's best guess, she can open the source and read the full context, including who wrote it.

Why this matters beyond onboarding

The same knowledge bank pays off well beyond the new-hire scenario. When an auditor asks why a particular adjustment was made two quarters ago, someone can search by topic or keyword and pull up the decision note in seconds, with author attribution intact — instead of reconstructing the reasoning from memory or old emails. When a vendor contract renews, the notes on past negotiations and pricing quirks are still there, versioned and editable, so nothing gets lost to a rewritten spreadsheet or a deleted email thread.

A few design choices make this realistic to actually use, rather than another tool that gets abandoned after the first month:

  • AI-drafted entries remove the main reason people never write things down: the blank page and the time it takes to make it presentable.
  • RAG question-answering means nobody has to know exactly which article to search for — they just ask, and get an answer with citations back to the source entries.
  • Categories, tags and full-text search make it possible to find things by topic, author or keyword, even months later.
  • Author attribution and view counts show who contributed and what's actually being read, so the knowledge base builds its own accountability.
  • Versioned editing lets anyone update an entry as a process changes, without wiping out the history of why it used to work differently.

Perhaps the most practical difference from a standalone wiki: this knowledge bank lives inside the same workspace as the company's books, customers and deals — not in a separate Notion account that someone has to remember to open. When the invoicing module, the CRM, and the process notes are all in one place, checking "how do we usually handle this" becomes part of the regular workflow instead of an extra app to log into.

The cheapest insurance policy

None of this requires a dedicated knowledge manager or a documentation sprint. It requires making the five-minute version of writing something down actually take five minutes. For a Swiss SME where headcount is tight and turnover is a when-not-if, a knowledge bank that people actually use is, as the saying goes, the cheapest insurance policy a small business can buy.

Stop Juggling Tools. Start Running Your Business.

Start your free account in 2 minutes. No credit card required.

Create Free Account