Before their CFO retired, Ryan McNulty, President at MBH Architects, made sure to get him in a room for a brain dump.

Over the course of a few conversations, Ryan and his team walked him through how he reviews contracts, specifically NDAs. Which clauses were always unacceptable. Which ones the firm would negotiate, and under what circumstances. What language a client should be expected to provide. What unusual clauses tend to appear, and how the firm has learned to respond. Decades of contract judgment, compressed over time into instinct, were suddenly made visible.

They distilled those conversations into a set of NDA best practices. Those best practices now power a contract review agent. Any manager at MBH can upload a new client NDA, and the agent goes clause by clause, marking each one high, medium, or low risk, explaining the rating, and suggesting a response. Expertise that once lived in one person’s head is now available to everyone in the firm, consistently, at any hour.

The technology mattered, of course. But it was not the only thing that made this possible.

What made the agent valuable was the judgment behind it. Someone had to understand that NDA review was important enough to improve. Someone had to know which expertise mattered. Someone had to help extract that expertise from the head of an experienced leader and turn it into guidance an agent could actually use. Someone had to see that this was more than an interesting AI experiment. It was a chance to take a recurring business process and make the firm’s best thinking more available, consistent, and scalable.

As we have been working with AEC firms in the Synthesis Knowledge Agent private beta, I have found myself coming back to a simple observation: the firms getting the most traction are having their firm leaders build agents themselves.

That distinction matters because high-impact knowledge agents are not created from AI enthusiasm alone. They are created when people who run AEC firms begin to personally and viscerally understand how agents can improve work that matters.

Over the past year, I've had countless conversations about AI agents across the AEC industry — with clients, partners, peers, and practitioners at every level. 

I keep noticing we're using the same word to describe wildly different systems.

Sometimes "agent" means an informational chatbot that can answer questions using firm knowledge. Sometimes it means a workflow assistant that helps generate proposals. Sometimes it means a system executing repeatable business processes with minimal human input. And sometimes it means fully autonomous systems coordinating multiple specialized agents across an entire operation.

That ambiguity matters. The way a firm should approach an informational agent is fundamentally different from how it would approach a fully autonomous system operating across multiple workflows. The technical requirements are different. The governance requirements are different. The trust, risk, and organizational implications are different. And yet most conversations about AI agents flatten all of this into a single category.

Autonomous vehicle companies solved a similar problem. Rather than debating whether a car was "autonomous" or not, they introduced a spectrum of capability levels — a shared language for discussing current capabilities, where they were headed, and what human oversight was still required along the way.

That framing inspired me. So I started sketching something similar for AEC: not a single definition of "agent," but a spectrum of increasingly capable systems — each with different strengths, risks, requirements, and use cases.

I've been calling it the Agent Capability Spectrum.

Imagine you’re early in your career and newly staffed on a healthcare project. You’ve been asked to help work through the onstage/offstage model for a new facility—a concept you’ve heard before, but you don’t feel fluent enough in to start designing with confidence. You know your firm has done this work many times, and you know the knowledge exists somewhere. What you don’t know is where to look or how to get started.

So you turn to your firm’s AI-powered Healthcare Knowledge Agent.

You ask a simple, situational question: What are the key considerations when designing an effective onstage/offstage model for a healthcare project? The response you get is grounded in how your firm approaches healthcare work. It draws from internal best practices, past projects, and recorded talks from senior healthcare leaders. It highlights what to pay attention to early, where teams often run into trouble, and how different decisions affect patient experience and staff workflows.

Along the way, it points you to specific internal resources—an upskilling video on healthcare planning, linked directly to the exact moment where this concept is discussed; a standards page that captures onstage/offstage best practices; and a case study from a past project that shows how these ideas come together in the real world. If you want to go deeper, you can. If you just need enough context to start shaping a design direction, you have it.

Later that week, you’re asked to begin contributing to an imaging suite—something you’ve never worked on before. You ask the Healthcare Knowledge Agent what’s important to consider. Again, the agent synthesizes the firm’s best thinking, pulling together insights from different people, in different formats, shared at different times. And when follow-up questions move into territory it can’t answer with confidence, it doesn’t pretend otherwise. It clearly signals the limits of its knowledge and points you to the right subject-matter experts to answer the hard questions.

You’re able to stay in the flow of work—learning as you go, making progress, and building confidence.

Now imagine the same moment from the other side of the equation.

You’re one of the firm’s healthcare leaders. You’ve spent years building deep expertise through projects, research, and mentoring. You care deeply about developing the next generation, but you also know how often your time is consumed by answering the same foundational questions—important questions, but repeatable ones. Questions that interrupt deep work and pull you away from clients, strategy, and the harder problems that really need your attention.

By contributing to your firm’s digital knowledge base—through interviews, recorded talks, and curated guidance—your expertise becomes accessible on demand, 24/7, in a form that’s contextual, searchable, and connected. When emerging professionals come to you with questions, they’re better informed and more specific. Your time is spent mentoring at the right level, serving clients, advancing the firm’s thinking, and continuing to add new insights that strengthen the firm’s collective knowledge over time.

In short, your expertise has been leveraged so you can have more impact and your firm’s emerging professionals can develop faster.

Once you can imagine this working in healthcare, it’s not hard to extrapolate the same pattern applying elsewhere—whether that’s a Sustainability Knowledge Agent helping teams navigate materials and approaches, or a Revit Knowledge Agent supporting designers while freeing design technology teams to focus on innovation. Across disciplines, the dynamic is the same: expertise scales, learning accelerates, and the organization gets smarter.

This vision explains why Knowledge Agents will be the next major pillar in the Synthesis platform, alongside Intranet, LMS, and AI Search capabilities.

In this issue of Smarter by Design, I’ll take you deeper into how Knowledge Agents will work, what processes and cultural habits will make them successful, and how you can begin to lay the foundation for their arrival later in 2026.

I found myself in a fascinating conversation with the COO of one of our clients last week. We were talking about something I’ve been circling around for months, but this discussion finally snapped the pieces into place.

It’s what I’m starting to call the AI and Expertise Paradox.

We all know the demographic story by now. Baby boomers are retiring in large numbers and there aren’t enough Gen Xers to replace them. 

In AEC, that often means we’re losing some of the deepest technical knowledge in our organizations—codes, construction standards, quality practices, the kind of judgment that only comes from decades of watching real projects go from concept to completion.

Technical experts possess the kind of deep smarts that can look at a drawing and feel that something isn’t quite right.

And they are retiring.

At the same time, we’re seeing a wave of AI-powered tools arrive that promise to help fill the gap. Automated code checks. QA/QC scanners. Plan reviewers that highlight potential issues a junior architect or engineer would never recognize. Assistants that allow someone to work across jurisdictions with different codes and standards and at least have a baseline level of support.

In some ways, it feels like knowledge augmentation—almost like the moment in The Matrix when the character Tank uploads the knowledge to fly a helicopter into Trinity’s brain.

Similarly, there are numerous emerging AI tools in our industry which, if they deliver on their vision, will enable a junior team member to run a basic code review, an expert who is stretched thin across multiple projects to offload routine checks, or an architect or engineer working on a project in a different region to get a helpful second set of eyes on local code compliance.

On the surface, this looks like the perfect solution: AI tools that allow those with less expertise or who are super busy to do more.

But here’s where the paradox emerges.