Knowledge Transfer Is a Format Problem

A few years ago, I sat in a meeting with the head of operations for a large engineering firm. They had a problem. One of their leading diagnostic engineers, a man who could tell you which specific bearing was about to fail on a multi-million-pound turbine just by listening to a sound file, was retiring in six months.

They wanted to ‘capture his knowledge’. Their plan was simple: get him to write it all down. A series of detailed guides, process documents, maybe some checklists. They’d film him talking, too. They would then give all this material to a junior e-learning designer to turn into a course.

I told them it wouldn’t work. Not because the engineer wasn't brilliant, or the designer wasn't competent, but because the entire premise was wrong. The format had already defeated them before they’d written a single word.

They were treating knowledge transfer as a logistics problem. How to get a payload of information from point A (the expert’s head) to point B (the company’s learning management system). But the most valuable part of that engineer's expertise wasn't a payload of facts. It was a finely tuned system of judgement. It was the messy, intuitive, experience-forged network of heuristics, pattern-matching, and sensory calibration that lived between the facts.

And you can’t capture that in a Word document.

The Misleading Simplicity of ‘Knowledge Transfer’

The phrase itself is the start of the problem. ‘Transfer’ suggests a clean, lossless process, like moving a file from one drive to another. The reality is more like making a photocopy of a watercolour painting using a machine that only sees in black and white. The shapes are there, but the life has gone out of it.

We do this all the time. We take a subject-matter expert, a person with decades of non-linear experience, and we force their knowledge through the most linear, flattening formats imaginable.

The Document Dump: The expert produces a 200-page manual. It's comprehensive, accurate, and utterly unteachable. No one has the time or motivation to read it. It sits on a server, technically 'captured' but functionally useless, a monument to a failed process.

The PowerPoint Autopsy: The expert’s slide deck, full of dense bullet points that acted as prompts for what they would say, is handed to a designer. The designer, tasked with turning it into a course, puts one bullet point on each screen and adds a ‘Next’ button. The result is a click-through presentation that tests recall of the bullet points, while the crucial context—the discussion, the caveats, the stories that brought the points to life—is lost forever.

The Talking Head: We film the expert. This is an improvement, as we at least capture tone and emphasis. But it’s still a broadcast. It positions the learner as a passive receptacle for information. Learning, especially the development of professional judgement, is not a passive activity. It is an act of construction, of argument, of trial and error.

The common thread here is a focus on the explicit – the facts, the rules, the processes. These are the easiest things for an expert to articulate and the easiest things for an organisation to package. But they often represent only a fraction of the real capability. The tacit knowledge – the ‘feel’ for a situation, the diagnostic sixth sense, the ability to see what isn’t there – is where the value lies. And these formats are fundamentally hostile to it.

Translation, Not Transcription

To preserve the signal, we have to stop thinking like archivists and start thinking like translators. The job of a good instructional designer, when faced with a true expert, is not to be a stenographer. It is to be a Socratic partner.

The goal is not to ask, “What do you know?” but to probe the structure of that knowing with better questions:

• “Talk me through the last three times you dealt with this. What was different each time?”
• “What’s the first thing you look for that a junior person always misses?”
• “When does that official process not apply? What do you do then?”
• “Describe a situation where the data looked normal but you knew something was wrong. What tipped you off?”
• “What’s the most expensive mistake you’ve ever seen someone make in this area?”

These questions don’t elicit lists of facts. They excavate stories, exceptions, and mental models. They make the expert’s invisible thinking visible. This is the central idea behind cognitive apprenticeship, a concept from the late 80s by Allan Collins and his colleagues that feels more relevant now than ever. The designer’s work is to surface the expert’s internal monologue and then build a sandbox where a novice can start to develop their own.

Once you’ve done that work of translation, you can choose a format that honours the complexity you’ve unearthed.

Formats That Build Judgement

If the goal is to build judgement, not just transmit information, then the learning format must require the learner to exercise judgement. It must give them a safe space to make decisions, see the consequences, and try again. For our retiring engineer, instead of a manual, we should be building things like this:

Scenario-Based Simulations: Present the learner with a realistic situation. “Here are the sensor readings from Turbine 7. Here’s the audio file. What’s your initial diagnosis? What single piece of additional data do you need to request?” Then, based on their choice, the scenario branches. The feedback doesn’t just say “Correct” or “Incorrect”. It says, “That’s a plausible first step, but a senior engineer would have also checked X, because in this model, Y is often a misleading indicator.”

Dialogue Systems: Instead of a text block explaining how to conduct a client interview, build a simulated conversation. Let the learner choose from three possible questions to ask next. Let the virtual client react accordingly. Show how one line of questioning closes down a conversation, while another opens it up.

Critical Reading Exercises: Give the learner a document — a draft report, a case file, a technical specification — and a task. “Review this as our expert would. Identify three hidden risks and one missed opportunity.” This forces them to apply the expert's lens, not just recall their conclusions.

These formats work because they don’t just present the knowledge; they embed it in the context where it is actually used. They shift the learner from being a spectator to being a participant.

Of course, building these rich, interactive formats used to be prohibitively expensive and time-consuming. It required specialist tools and teams. That’s changing. Modern authoring platforms, including the one I’ve been building, dramatically lower the barrier to creating this kind of learning. The AI can handle the heavy lifting of structuring a scenario or building out a decision tree, freeing up the designer to focus on the most important part: the quality of the questions and the subtlety of the feedback.

The technology makes it faster, but it doesn’t replace the critical act of human-to-human translation. You still need someone to ask the right questions of the expert, to find the hidden gold. But now, once you have it, you can build a fitting vessel for it in hours, not months.

We can’t afford to keep letting deep expertise evaporate on the winds of retirement or get flattened into useless PDFs. The cost is too high.

The solution isn't to ask our experts to work harder at documenting what they know; it's for us to get smarter about translating how they think.