The owner of an eighteen-person consultancy opened a folder on a Friday afternoon and tried to work out what would happen when her operations lead retired in six months. He had been there since year three. She had a procedures handbook he had written in 2022, a shared drive of templates, and a project tracker. None of it told her what she actually needed to know. It did not say which clients he rings rather than emails when a deadline is slipping. It did not say which combination of project signals makes him quietly add a fortnight to the timeline before anyone notices. It did not say how he knows, usually within an hour, that a particular supplier is about to miss a deliverable. The handbook described the process. The patterns lived in his head.
That is tacit knowledge, and it is the part of an owner-operated business that walks out the door with experienced people unless you capture it deliberately. One Walden University doctoral study estimates it makes up around 85 per cent of what a senior person carries in their head. The cost of losing it shows up as months of pattern-matching the next person cannot yet do, not as a missing handover document.
What is tacit knowledge and what gets lost when it leaves?
Tacit knowledge is the judgment, intuition and pattern recognition someone builds up by doing the work for years. AIIM frames it as the opposite of explicit knowledge, which lives in policies, standard operating procedures and training material. Explicit knowledge survives a handover because it has already been written down. Tacit knowledge does not, because the person holding it usually cannot fully describe what they do without prompting. They just do it.
In an owner-operated firm, the people who hold the most of it are the long-serving operations lead, the senior delivery manager, the account director who has carried the largest client for a decade. Walden’s case research on insurance agency knowledge loss found that when those people leave, replacement productivity takes between eight and twenty-six weeks to recover, depending on the complexity of the role. The handover documents do not shorten that curve. The patterns do, and the patterns are exactly what the documents miss.
Why do documentation projects miss what matters?
Many owners respond to a looming retirement by commissioning a documentation project. A new hire or consultant writes down how things work, producing a folder that looks complete. Six months later, with the experienced person gone, the real gaps show up. The Project Management Institute’s research on tacit knowledge in projects found exactly this pattern even in environments where documentation discipline is strongest.
Documentation projects capture what people consciously know they know. They miss what people unconsciously do. An experienced account director can write down the client reporting process because she does it deliberately. She does not write down that she always checks one specific line of a particular client’s report three times because a rounding error once led to a complaint she remembers vividly. The why is what travels. The what, on its own, does not.
The second failure mode is replacing conversation with form-filling. Alchemy Solutions’ work on knowledge transfer in established businesses argues that knowledge evolves through dialogue, and a forty-page document read in isolation cannot replicate the back-and-forth that surfaces judgment. The replacement absorbs the steps and misses the reasoning. When circumstances change, as they always do, they have no idea which rules to bend.
What capture methods actually work at SME scale?
Three methods work without enterprise platforms or external consultants, and a five-to-fifty-person firm can run all three with existing tools. Recorded walk-throughs capture knowledge in motion, paired observation captures it through proximity, and structured interviewing on real cases captures it through prepared conversation. Used together they produce a richer record than any documentation project, and the production quality matters less than the clarity of the narration.
Recorded walk-throughs are the simplest. The expert performs a real piece of work, narrates what they are doing and why, and records the screen with a phone or built-in Zoom or Teams capture. Short is better than long. A ten-minute walk-through of an actual refund exception is more useful than a forty-page document, particularly when the recording is transcribed into searchable text. The expert is least likely to forget edge cases when they are actively in the work.
Paired observation is older and labour-heavier. A newer team member sits alongside the expert for two to four weeks on a specific process area, takes notes, asks why repeatedly, and runs a fifteen-minute debrief at the end of each day. BambooHR’s research on job shadowing is clear that the structure is what makes the difference. Without it the observer drifts and the expert feels interrupted. With it, the observer leaves the period able to recognise the patterns the expert recognises and ask informed questions on the ones they cannot yet see.
Structured interviewing on real cases is the third leg. Enterprise Knowledge’s practical guide describes the protocol: identify three to five real cases the expert has handled, then run three or four forty-five to sixty-minute sessions over a month asking what they noticed first, what options they considered, why they chose the approach they did, and what could go wrong with this type of situation. The interviewer does not need domain expertise. They need to ask why repeatedly and resist filling silences. A manager, a peer or an external facilitator can run the sessions effectively from a prepared script.
What can AI actually do with the knowledge once you have it?
AI is genuinely useful with captured tacit knowledge, but only after the capture work is done. It transcribes recordings into searchable text, so a new hire can ask a question and land on the exact moment in an interview where the expert answered it. It summarises long walk-throughs into key decisions, groups patterns across interviews, and flags gaps in what has been captured so far.
What AI cannot do is exercise original judgment on a novel situation. Harvard Business School’s recent work on AI and human judgment found that generative AI produces generic suggestions on strategic decisions, the kind a less experienced person could also generate, and that the real value comes from experienced operators critically evaluating AI output rather than relying on it. KMS Lighthouse makes a related point about data quality risk: AI trained on incomplete or biased captured knowledge will amplify the gaps, recommending only the standard approach because the unconventional cases were never recorded. AI is a force multiplier for captured knowledge. It is not a substitute for the capture discipline.
What does a sustainable capture cadence look like?
The failure mode owners fall into is treating capture as a six-week project triggered by a retirement announcement. Six weeks of panic, then nothing. By the time the replacement is fully embedded, new edge cases the project never reached have already emerged. Joysuite’s operational work describes the alternative: fifteen minutes a week per knowledge holder, built into normal work rhythm and continued after the expert has gone.
Fifteen minutes a week sounds slight and is not. Over a year it compounds into roughly thirteen hours of recorded knowledge per person, built from real work rather than retrospective reflection. The lightweight template is three questions: what did I just do, what was the tricky bit, what do I want someone else to know. Answering them into a voice memo is far more likely to happen than writing a polished two-page document. The owner’s job is to signal that this matters, model it personally, and create enough psychological safety that the expert is open about what they do. Harvard Business Review’s work on why employees do not share knowledge makes the safety point sharply: people will not be fully open in interviews or walk-throughs if they feel they are documenting themselves out of a job. The owner has to make clear that the purpose is keeping the business running, not engineering anyone out of a role.
If you want help working out what a proportionate capture discipline looks like in your firm, book a conversation.
