---

I found myself in a leadership conversation a few weeks ago talking about mentorship, and I gave the same answer I think a lot of us give.

“It’s part of our culture.”

After I said it, I realized how weak that answer actually was though.

How many junior staff in most firms actually have an active mentor? How are they paired? What happens when that mentor leaves? How do we know if any of it is even working?

Truthfully, most AEC firms don’t really know. We kind of eyeball it, hope good relationships form naturally, and when someone resigns, the mentorship usually walks out the door with them.

That’s not a mentorship strategy. That’s just hoping things work themselves out.

The Silent Crisis in AEC People Ops

A few months before that conversation, I watched a senior designer put in their notice. Obviously, there are project impacts when that happens. Schedules shift, teams scramble, and risk goes up. But what stuck with me afterward wasn’t just the delivery side of it. It was realizing that this person had also been mentoring several younger staff members every day in ways nobody was formally tracking. Helping them grow, answering questions, guiding decisions, teaching them how to navigate the work and the industry. And the day they left, all of that disappeared too.

Nobody really talks about that side of turnover.

Mentorship doesn’t get treated with the same level of operational attention as project delivery. Even firms that genuinely care about staff development usually don’t have much structure around it. When senior staff leave, there’s rarely a transition plan for the mentorship relationships they carried. No visibility, reassignment, or continuity.

In project delivery, we’d never tolerate that. If a PM left in the middle of a major project, leadership would immediately step in and reassign responsibilities. But when mentorship disappears? Usually…silence.

And honestly, this problem is only getting bigger. Mid-career tenure is shrinking. Senior bench strength is getting thinner. We’re bringing in plenty of new graduates, but we’re not growing the mentor capacity needed to support them at the same pace. We’re heading toward a point where the demand for guidance is going to exceed the number of humans available to provide it.

The Hybrid Mentor

Alright, full warning here… yes, this is where AI enters the conversation. But stick with me.

For years, we’ve treated mentorship as something that could only exist as a one-to-one human relationship. And obviously that human component still matters tremendously. But technology has finally matured enough that I think there’s a second layer we should be thinking about. Not AI replacing mentors, but AI supporting them.

Imagine a junior engineer paired with a senior leader for the relationship side of mentorship providing judgment calls, career advice, navigating difficult situations, leadership growth, and all the human stuff that actually matters. Then underneath that, they also have an AI mentor layer handling the day-to-day support like workflow walkthroughs, standards questions, “how did we solve this on the last project,” tool guidance, process navigation, and internal best practices. The kind of things junior staff need constantly, but often hesitate to ask because everyone around them already seems overloaded.

That changes the equation completely.

The senior mentor stays focused on the high-value conversations only humans can really provide. The junior staff member gets faster answers, more confidence, and more support exactly when they need it. And maybe most importantly, the knowledge itself stops living exclusively inside one person’s head. It becomes part of an accessible system that doesn’t retire, resign, or get reassigned to another project.

Where The Knowledge Actually Comes From

This is usually the first thing people ask me when I bring this up: “What would the AI even learn from?”

Honestly, most firms are already sitting on massive amounts of usable knowledge. They just aren’t leveraging it well.

Every AEC firm already has years of standards, templates, lessons learned, workflow guides, and “how we actually do things at this company” types of documentation. The problem is most of it is buried somewhere in SharePoint or a server nobody opens unless they absolutely have to. An AI layer suddenly makes all of that searchable and conversational. That standards document nobody reads now becomes the standards conversation people can actually interact with.

Then there’s the vendor ecosystem side of it. Companies like Autodesk, Bentley Systems, and Esri already publish enormous libraries of technical documentation, workflows, implementation guidance, best practices, and forums. That’s an incredibly deep knowledge base most firms still aren’t connecting into their AI ecosystems. When you combine vendor knowledge with your own internal standards, you suddenly create a system that understands both how the tools are intended to work and how your firm specifically uses them.

But honestly, the biggest knowledge source is probably the one almost nobody is talking about.

Your employees are already generating mentor-level knowledge every single day through AI conversations.

Think about it. A designer using Claude to troubleshoot a Civil 3D corridor issue at 11 PM. A PM using ChatGPT to structure a difficult proposal response. A junior engineer walking through a workflow problem during lunch with Gemini. Someone refining a drainage calculation with AI support. Every one of those conversations produces useful, contextual, real-world AEC knowledge.

And right now, most of it disappears forever inside private chats.

That’s one of the biggest silent knowledge leaks happening in our industry right now. Firms are paying for AI tools. Employees are getting real value from them individually. But the organization itself isn’t capturing any of the institutional learning being created in the process.

Of course, none of this works unless employees trust the system.

How To Do This Without Burning Trust

The firms that figure out how to ethically aggregate and curate that internal AI usage are going to build mentor agents that are genuinely useful instead of generically helpful. Your own people are the best curators of what good looks like inside your firm. You just have to give them a way to contribute, and a reason to.

First, move your AI tooling from personal accounts to enterprise seats. The enterprise versions of ChatGPT, Claude, and Gemini now include data agreements where your conversations stay inside your firm, don’t train the vendor’s external models, and can be governed centrally. That’s the foundation. Without it, you have no legitimate path to aggregation in the first place.

Second, build opt-in contribution into the workflow. Give employees a simple way to flag a specific exchange and say “this one is worth saving for the firm.” Could be a button, a tag, a shared channel…whatever fits your stack. The employee retains control, can review what they’re contributing, can redact what they want redacted, and gets visibility on what gets used. Nothing leaves their account without their decision. That last part matters because the fastest way to kill participation is to make people feel surveilled.

Third, set up a curation layer. Treat AI knowledge contributions the way you’d treat any other technical content. A small group of senior engineers or digital delivery staff reviews submissions, strips identifying details, validates accuracy, and decides what gets promoted into the mentor agent’s training set. The same discipline as you apply to standards updates coupled with the same governance you apply to BIM content.

None Of This Works Without Measurement

Even with great mentors and great AI systems, if you can’t actually measure mentorship coverage, you’re still operating on hope.

Mentorship needs the same level of rigor we already apply to utilization, BIM governance, QA/QC, and project delivery. Firms should understand who has mentorship support, where the gaps are, whether relationships are continuing over time, and whether the system is actually helping people develop. Because what leadership measures is usually what leadership prioritizes.

This is a big part of what I’ve been thinking about while building an intelligent project resource optimizer myself, so that it creates visibility into mentorship coverage, pairing support, and giving leadership a real operational view of workforce development instead of relying on assumptions and good intentions.

At the end of the day, mentorship without infrastructure is just hope. Mentorship with infrastructure becomes leverage.

The firms that figure out how to combine human mentorship, AI support layers, institutional knowledge capture, and measurable workforce development into a single ecosystem will certainly have an advantage. The firms that develop people the fastest are typically the firms that outperform everyone else.

How is your firm handling mentorship today? And honestly…where is all of your AI-generated knowledge living right now?