Arusto

Most organizations think they have a knowledge management problem. They don’t. They have a content creation problem. The distinction is costing them more than they realize.

At Arusto, where we build an AI creation layer for enterprise learning, we see this every day: organizations watching expertise disappear faster than they can turn it into something others can learn from. Sometimes it’s a retirement. Sometimes it’s an employee leaving for a competitor. Sometimes it’s a restructuring that moves people between teams. The trigger varies. The result is the same.

---

The thing documentation doesn’t capture..

A few years ago, I watched a large organization lose one of its most valuable employees. Not to a competitor. She retired.

They had documentation, process guides, training materials, recorded knowledge transfer sessions. Within weeks, people were asking questions nobody could answer. Why was a critical system designed that way? What had been tried before? Which shortcuts were safe and which ones caused failures six months down the line?

The information existed, but it had never been written down, because most of it wasn’t the kind of thing you write down. It lived in the judgment and pattern recognition of someone who had spent decades solving the same problems in slightly different forms. When she left, that knowledge left with her.

NASA learned a version of this lesson after the Apollo era. Years of budget pressure encouraged experienced engineers to retire. The technical documents still existed. The procedures still existed. The facilities still existed. But researchers later described what followed as a “hollowing out” of institutional knowledge. Decades of practical experience had disappeared with the people who built the systems.

Consider the maintenance supervisor who can tell a machine is about to fail because it sounds slightly different during a morning inspection. There’s no database field for that. No checklist item that fully captures it. It’s twenty years of accumulated pattern recognition compressed into a five-second call. The same thing happens with engineers who understand why a system was architected a certain way, or compliance specialists who know which risks deserve immediate attention.

Panopto’s Workplace Knowledge and Productivity Report found that 42% of institutional knowledge exists only in employees’ heads. The average large U.S. company loses roughly $47 million in productivity each year because of inefficient knowledge sharing. Knowledge workers spend more than five hours every week either searching for information or recreating knowledge that already exists.

More than 11,000 Americans reached retirement age every day. The knowledge management software market is projected to reach $16.2 billion in 2026. Nearly 60% of enterprises plan to increase spending on knowledge management. These tools are good at what they do. They improve search, retrieval, and access to information that already exists in a usable format.

But that’s the flaw. Most knowledge management platforms assume the knowledge has already been converted into something learnable. They organize it well, but they don’t create it.

Why the standard response keeps failing

The traditional knowledge transfer process looks roughly like this: a subject matter expert announces they’re leaving, someone schedules interviews, an instructional designer takes notes, a few recordings are made, documentation gets assembled, and eventually something appears in the LMS. If everything goes well, the process takes several months. Retirement timelines don’t wait for content production timelines.

More fundamentally, tacit knowledge doesn’t transfer well through interviews and documentation even when there’s time. Procedures, workflows, and policies can all be documented. What’s much harder to document is judgment. Why did you make that decision? What warning signs were you looking for? What does a problem look like before it becomes obvious to everyone else?

And the senior engineers who understand why a system was architected a certain way aren’t refusing to document their knowledge because they don’t care. They’re usually the busiest people in the organization. Asking them to spend ten additional hours every week documenting decades of experience is asking them to take on a second job while continuing to do the first one.

So the knowledge stays locked inside their heads.

What actually solves this

The bottleneck isn’t storing knowledge. It’s converting expertise into something another person can actually learn from, before it disappears. Let me walk through what that looks like when it works.

The capture process itself doesn’t fundamentally change. The expert still sits down for an interview. They still record themselves on a webcam, or host a Zoom session, or put out voice notes alongside their procedural documents. What changes is that they don’t have to think about how to structure any of it.

If we ask the right questions, if we have the right storylines and scenarios in mind, the expert can literally just talk. A conversational data dump. No formatting, no slide decks, no worrying about whether this is organized well enough for someone else to follow. Just decades of experience, spoken out loud.

Now, anyone who has ever watched an unedited knowledge transfer recording knows what this raw material looks like. People repeat themselves. They go on tangents. They over-emphasize their own role in a project. They circle back to the same point three different ways. There are speech fillers, anecdotal detours, and long stretches where the important insight is buried inside ten minutes of context-setting. These are all natural human tendencies. Nobody finds it useful to watch forty-five minutes of unstructured, repetitive video by someone who is thinking out loud.

This is where the content creation layer matters. Not the capture. The conversion.

An AI creation layer takes that raw conversational dump and does several things simultaneously. It identifies the core decision-making frameworks the expert described and separates them from the noise. It clips the irrelevant sections. It structures the remaining content into consumable segments. It builds visualizations automatically based on what the expert described, because when that person was talking, they had scenarios in their head, they had visualized information flows, they had seen systems working and failing. Putting an avatar of their face with some bullet points on a slide deck doesn’t capture any of that.

The system also handles the things organizations worry about. Content gets segmented by role, so the training a frontline technician sees is different from what a mid-level manager gets. Sensitive information is filtered. Compliance requirements are applied. Nothing non-compliant or risky reaches employees, but the process doesn’t become a bureaucratic bottleneck to get there.

The final product isn’t a single recording with annotations. It’s multiple small assets that learners can consume in whatever format works for them. Scenarios built from the stories the expert told. The expert’s voice and perspective preserved, alongside the organization’s guidelines and standards. Proper visualization, proper storytelling, proper production quality. It feels like high-quality learning content, and it came from someone sitting in a chair and talking for an hour.

Why this changes more than the offboarding problem

If the process of converting expertise into learning content becomes this straightforward, why would you wait until someone leaves to do it?

Imagine doing this every quarter. Your best people sit down for a 30-minute conversation about what they’ve learned in the last three months. New techniques they’ve developed, problems they’ve solved, patterns they’ve noticed. The system compiles it, structures it, and adds it to the organization’s knowledge base without anyone spending weeks on production.

Or every month. Or every week. Imagine capturing knowledge on the go and letting the system compile it once it has enough material down the line. You’re not asking anyone to do extra work. You’re not hiring instructional designers to shadow every team. You’re just letting the people who know things talk about what they know, and letting the system handle everything between that conversation and a finished learning asset.

The knowledge loss problem is real and urgent. But the bigger opportunity is building a system where institutional knowledge accumulates continuously, without it being burdensome. Where the organization’s collective expertise isn’t locked inside people’s heads waiting for an exit interview to come out. Where knowledge capture is just something that happens as part of how work gets done.

The organizations that figure this out won’t just survive the next wave of departures. They’ll have built something their competitors don’t have: a living, growing repository of institutional expertise that gets better every week, because the hardest part of the process, turning raw knowledge into something someone else can actually learn from, no longer takes months of manual effort.

That is the gap most organizations still need to close. And for many, the window is closing faster than they’ve planned for.