Knowledge transfer: how do you keep the expertise of people who leave?

What does an employee take with them when they leave?

The files stay on the server. What leaves is knowing how to use them: where the right version is, which old case looks like the one that just came in, why a given option was turned down for a given client.

The philosopher Michael Polanyi put it this way in 1966: we can know more than we can tell. This is what's called tacit knowledge. An expert doesn't think to pass on what seems obvious to them, and their successor doesn't know what to ask until the case comes up.

A retirement can be planned. A resignation leaves a few weeks' notice, and a long sick leave leaves none. In every case, the knowledge lived in someone's head rather than somewhere you could search.

What are the usual ways to transfer knowledge?

Most companies combine several of these:

• Mentoring or pairing: the person leaving works with their successor for several months, on real cases.
• Documentation: procedures, job descriptions, an internal wiki, a handover file.
• Handover interviews: someone questions the expert about their cases, contacts and methods, and writes down what they say.
• Knowledge mapping: spotting who is the only one with a given skill, so you can deal with it before they leave.
• Communities of practice: regular meetings between people in the same role to share hard cases.

Why don't these methods always work?

They take time, just when there is least of it. The person leaving is wrapping up their work, the successor is taking on theirs, and the handover happens in the last few weeks.

What gets written down is incomplete and goes stale fast. A forty-page handover file gets read once and never opened again: six months later, nobody remembers the answer was in there.

Above all, the useful question comes up after the person has left. It's when you open an old client's file a year later that you wonder what was promised and why. No handover interview could have seen that question coming.

What does an AI agent connected to the company's documents change?

Much of that know-how left a trail: proposals, reports, meeting notes, memos, correspondence with clients. An AI agent reads those documents where they are, in SharePoint, OneDrive or Dropbox, and links them to one another by client, project, method and team.

The successor can then ask the question when it comes up, in plain language: "What did we propose to this client in 2023, and why not the other option?" The agent answers and cites the documents behind its answer, so you can check.

Expertise no longer depends on one person's memory or on a handover file someone has to remember to reread. The idea is explained on the page about the AI context layer, and the search itself on the one about AI enterprise search.

What can you still ask after someone has left?

A few examples of what teams ask an agent connected to their files:

• Who has worked for this client before, and on what kind of project?
• What method did we use on a similar engagement, and what deliverable came out of it?
• How did we answer this objection in the last bid of this kind?
• Why did this project fall behind, according to the meeting notes at the time?
• Which version of the procedure applies, and what changed from the previous one?
• What commitments were made in this client's contract?

How do you prepare before someone leaves?

The two approaches work together. In order:

• Identify the knowledge at risk: planned departures, and people who are the only ones who know a client, a product or a method.
• Check where their trail lives: shared folders, email, business software. There's no need to tidy it up first.
• Connect an agent to those sources while the expert is still there.
• Have them test the agent on their own files: they'll see right away what's missing or misread.
• Spend interview time on the "why" that isn't written anywhere, and file those notes with the other documents so the agent reads them too.
• Keep the mentoring for hands-on skills and client relationships, which documents don't pass on.

What can't an agent replace?

A hands-on technique, a client's trust, the instinct that comes from twenty years in the trade: those pass on by working next to someone. The agent keeps what left a written trail and makes it easy to find. It doesn't replace mentoring, or the judgment of the person taking over the file.

What was never written down can't be found. That's why it pays to get the expert talking before they leave, and to put those notes in the files.

What does Saména do?

Saména builds custom AI agents for small and midsize companies. They read your documents read-only, in the folders you choose on SharePoint, OneDrive or Dropbox, and link each document to the others by client, project, method and team. Other tools, such as email, can be connected on request.

Your team asks its questions in the Saména app, and also from Claude, ChatGPT or Copilot. The agent cites its sources and can draft a Word document or a PowerPoint deck from what it found.

Your files stay in your folders. What Saména retains from them is encrypted with a key unique to your company, on servers in France with the database in the European Union. You choose the model: Claude, models from OpenAI, or Mistral if you prefer a French model. They run in Europe, with no retention of your data.

It all starts with a free 30-minute call, with no commitment. Two weeks later, a first agent is running.

Frequently asked

How do you pass on the know-how of an employee who's retiring?

Combine mentoring on real cases, interviews focused on the "why" behind decisions, and an AI agent connected to the company's documents, which the expert tests before leaving. Questions that come up after they've gone will then get an answer with sources.

What is tacit knowledge?

What a person knows how to do without being able to explain it: instinct, habits, the reasons behind a choice. It doesn't pass on well in writing. Part of it still lives in the documents produced along the way, and that's the part an agent makes searchable.

Do the documents need to be organized before connecting an agent?

No. The agent reads folders as they are and links documents to one another by client, project, method and team.

Can an agent read the email of someone who's leaving?

Email can be connected on request. Before doing so, check with your data protection officer what the GDPR and employment law allow, especially informing employees and leaving personal messages out.

Does the knowledge stay with the company if you change vendors?

The documents do: they stay in your folders, and the agent only reads them. What the vendor retained from them should be deletable; at Saména, it's deleted on request.

How long before a departure should you start?

As early as possible, so the person leaving can test the agent on their own files. At Saména, a first agent is running two weeks after the call.