Every engineering firm has one. The person everyone walks over to when the problem is actually hard. They know why the detail is the way it is, which supplier burned you back in 2011, and the failure nobody likes to talk about that still shapes how they design today. None of it is in a document. It’s in their head.
And in a few years, they retire.
That’s the part that should bother you. Most of the risks you manage are maybes. This one isn’t. It’s a date. When they walk out, decades of knowledge walk out with them, and you won’t feel it until the day someone needs an answer that used to take five minutes and now takes a week.
This isn’t a niche problem. Around 10,000 people hit retirement age every day in this country, and a big share of the experienced engineering and manufacturing workforce is on its way out over the next decade. The firms that get hit hardest are the ones where nobody ever wrote anything down, because they never had to. The expert was always right down the hall.
The tools are catching up
A handful of AI products have hit the market to take this on, and they come in two flavors.
The first kind watches the work and pulls knowledge out of it quietly, in the background. CoLab does this inside engineering design reviews, building a record out of the comments people leave on drawings and models. Others, like eGain, amaiko, and the search platform Glean, do the same thing across Teams, Slack, SharePoint, and email. Nobody has to do anything extra, which is the whole pitch.
The second kind goes and asks. Instead of waiting for knowledge to show up, these tools sit your people down and interview them. Morphe, a newly launched Canadian product, runs an employee through a few short AI-led interview sessions and turns the answers into something the rest of the team can use. A couple of others, Interloom and Cloneable, focus on feeding that captured expertise into AI agents, and they’ve raised real money doing it.
It’s good progress. Three years ago none of this was realistic. Now it is. But before you put any of it in a budget, know what they don’t do.
Where they come up short
They mostly capture going forward, not backward. The watch-the-work tools only catch what flows through them after you turn them on. They do nothing about the 25 years already in someone’s head, and nothing at all if that person is gone next quarter. If your problem has a date on it, a tool that slowly fills up over years is solving a different problem.
They’re built for operations, not engineering. Most of them are general-purpose, or aimed at field service and customer support. They nail “here’s the workaround” or “here’s who to call.” They’re not built to capture why a design went the way it did, what failed last time, or how a 30-year engineer reads a spec that isn’t black and white. In your shop, that’s the part that’s actually worth money, and it’s the part these tools handle worst.
They’re sized for big companies. The serious platforms come with six-figure price tags, large seat minimums, and rollouts measured in months. If you’re a mid-sized firm trying to capture one or two irreplaceable people before a known retirement date, that’s the wrong size and the wrong clock. The platform goes live, and your expert is already gone.
And the easy ones stay shallow. Here’s the catch nobody mentions. The tools that capture without any effort tend to only catch the surface. The knowledge that actually matters, the judgment your expert can barely put into words, only comes out when somebody deliberately pulls it out of them. That takes effort, which is exactly what the effortless tools are built to avoid.
What you can actually do right now
Here’s the good news, and it’s better than the sales pitches let on. Capturing what one retiring expert knows is a process problem, not a software problem. You can start this quarter for about the cost of a few software seats.
It’s not complicated. Pick the three to five people whose leaving would hurt the most, weighing how soon they’re out against how hard they’d be to replace. Sit them down for a handful of recorded conversations. People will say in an hour what they’d never sit down and type. Use a regular AI tool to build the questions ahead of time, then to turn the recordings into something organized and searchable afterward: the decisions, the reasoning, the war stories, the people they lean on. Then make sure the next engineer can actually find it.
The one part you can’t cut corners on is the questions. For an engineering firm, you have to dig at the stuff that matters: why a design went one way and not another, what has gone wrong in the field and why, how your best people read the standards that leave room for judgment, which fabricators and suppliers they trust and for what. Ask generic questions, get generic answers. That’s the real work, and you can start it without waiting on a single purchase order.
Buy a platform later if you decide you want one. Just don’t let shopping for it hold up the capture. The platform doesn’t have a deadline. Your people do.
The bottom line
The firms that come out of the next ten years in good shape won’t be the ones with the biggest knowledge-management system. They’ll be the ones who treated their experts’ retirements like the scheduled events they are, and got the knowledge down while there was still somebody around to ask.
The technology is finally good enough to help. The only thing you can’t buy more of is time, and for some of your best people, the clock’s already running.
