Capture Tacit Knowledge With AI Before Key Staff Leave
Most SME know-how lives in a few heads. Use AI interviews, meeting notes and screen recordings to turn it into a company second brain before people leave.
Why tacit knowledge is an SME's biggest hidden risk
In most small businesses, the knowledge that keeps things running is tacit: it lives in the heads of the owner and one or two long-serving staff, not in any document. When one of them resigns, retires or goes on long leave, work slows down, mistakes creep in and clients notice.
You usually discover this only when you try to write a process down. Things run smoothly because someone is quietly holding everything together, as we noted in Does your business actually need AI? An AI second brain can fix this, but only once the knowledge is captured. This playbook shows how to capture it with AI doing most of the writing. For the bigger picture, see our pillar guide to AI second brains.
Step 1: Map who knows what
Start by listing the people whose absence would hurt most, and what only they know. Spend 30 minutes with the owner or managers on three questions:
• If this person were away for a month, which tasks would stall?
• Which questions do other staff always bring to them?
• Which client relationships, supplier arrangements or workarounds exist only in their head?
You'll end up with a short list of people and topics. Prioritise by risk: someone nearing retirement or likely to leave comes first, even if their topics seem small.
Step 2: Run AI-assisted knowledge interviews
The fastest way to capture tacit knowledge is a recorded conversation that AI turns into a draft document. Experts rarely have time to write, but most will happily talk for 30 minutes.
How to run one:
• Pick one topic per session, such as "how we quote for custom orders".
• Record it with a meeting-notes tool or your phone, with the person's agreement.
• Ask for real examples: "Walk me through the last time you did this." "What went wrong once, and what did you change?" "When do you break the rule?"
• Have AI draft the document: paste the transcript into Claude, ChatGPT or Gemini and ask for a step-by-step SOP with a "common mistakes" and "exceptions" section.
• Send the draft back to the expert to correct. Editing a draft takes them ten minutes. Writing from scratch would take an afternoon and probably never happen.
The exceptions section matters most. That's where tacit knowledge hides: the client who always needs extra lead time, or the supplier whose invoices need checking.
Step 3: Capture how the work is done on screen
For software-heavy tasks, a narrated screen recording captures more than any written description. Ask the expert to record themselves doing the task once while explaining what they're doing and why.
Then use AI to turn it into a written guide. Many AI tools can now summarise a video or its transcript into numbered steps. Keep the recording as a reference alongside the written steps. Written steps are what the second brain searches, and the video is there for anyone who needs to see it done.
Step 4: Make meetings and decisions capture themselves
The cheapest knowledge to capture is the knowledge created every week in meetings and chats, so capture it as it happens. Most decisions in an SME are made in a meeting, a WhatsApp thread or a quick call, and then forgotten.
Set up simple habits:
• Use AI meeting notes for internal meetings and client calls where participants agree to it, and save summaries to one shared folder.
• Keep a decision log: one line per decision with the date, what was decided, why and who decided. AI can extract these from meeting summaries automatically.
• Move important chat decisions into a document. If it was agreed on WhatsApp, someone copies it into the client file.
Over months, this becomes the history your second brain draws on to answer "what did we agree with this client, and why?"
Step 5: Structure it so the AI can use it
AI answers best from short, clearly titled documents with one topic each. Turn the drafts from steps 2 to 4 into a consistent format:
• A clear title that matches how people ask: "How to quote a custom order", not "Quoting v3 final".
• Owner and "last reviewed" date at the top.
• Steps, then exceptions, then common mistakes.
• Links to related documents.
Store them where your second brain can reach them, then connect them to your AI tool. Our step-by-step build guide covers the setup, and the tools comparison helps you pick the platform.
Step 6: Handle the people side with care
Knowledge capture only works if experts feel valued rather than replaced. A long-serving employee asked to "write down everything you know" may reasonably wonder whether they're being prepared for the exit.
Be explicit about the reason: the business wants to reduce the load on them, so they stop being interrupted and can take proper leave. Credit them as the owner of their documents. Frame it as their expertise becoming the company standard.
If staff are anxious about AI more broadly, our guide on introducing AI to your team without the job-loss fear covers that conversation. Get consent before recording anyone, and keep personal data out of process documents, in line with our PDPA and AI second brain checklist.
A realistic 60-day plan
Most SMEs can capture their riskiest knowledge in about two months with a few hours a week.
• Weeks 1–2: map who knows what; choose the top five topics.
• Weeks 3–6: run one or two 30-minute interviews a week; turn each into an AI-drafted SOP; have the expert correct it.
• Weeks 5–8: switch on AI meeting notes and a decision log; add screen recordings for software tasks.
• Week 8: connect the documents to your second brain and test it with the questions staff actually ask.
Then keep going at a slower pace: one new topic a fortnight, plus a monthly review. If you'd like help getting started, our AI Foundations training covers these capture habits on your own tools, or book a free consultation.
Frequently asked questions
What is tacit knowledge in a business?
Tacit knowledge is know-how people carry in their heads rather than in documents: how to handle a difficult client, which supplier needs extra checks, when to bend a rule. It's built from experience and is hard to write down, which is why it's lost when experienced staff leave. AI helps by turning conversations into drafts.
How can AI help capture employee knowledge?
AI can turn a recorded 30-minute conversation or a narrated screen recording into a structured draft SOP in minutes. The expert then corrects the draft rather than writing from scratch. AI meeting-notes tools can also capture decisions automatically, so new knowledge is saved as it's created instead of being lost in chats.
How do I get staff to share their knowledge?
Explain that the goal is to reduce their load and interruptions, not to replace them. Keep sessions short, ask for real examples, and let AI do the writing so they only review. Credit them as owners of their documents, and get their consent before recording. People share more readily when it saves them time.
Is it legal to record staff for knowledge capture in Singapore?
Recording with the person's knowledge and agreement for a clear business purpose is generally fine, but personal data in recordings is covered by the PDPA. Tell staff what's recorded and why, store recordings securely, limit access, and delete them when no longer needed. Keep personal details out of the resulting process documents.
