In this episode of the Smarter by Design podcast, I’m joined by Nicole Chavas, President and COO of Greenprint Partners, and Hillary Thompson, Principal at MBH Architects, for a candid conversation about what they are learning as they turn AI experiments into meaningful business transformation.
Both Nicole and Hillary are senior leaders who are also personally building agents and redesigning workflows, which gives them a close-up view of not only what the technology can do, but what it reveals about their firms. Again and again, they have found that building with AI exposes the same underlying questions: What does good look like? Is this actually our standard, or just what we happen to do today? Is the knowledge documented clearly enough to use? And are we redesigning the work itself, or simply bolting AI onto an existing process?
One of the clearest lessons is that scalable transformation often means going smaller, slower, and together. Nicole describes initially trying to move quickly, work largely on her own, and use agents to redesign entire business processes at once. What she learned was that proposal development and project management are really collections of interconnected subprocesses, and that lasting change requires a different approach: involve the people doing the work, break the system into smaller pieces, focus on the highest-impact opportunities, define what good looks like, redesign each subprocess for an AI-enabled way of working, and then train the firm on the new process.
Hillary describes a parallel journey at MBH, where rapid prototyping with firm leaders through an “agent charrette” surfaced her leadership team’s dreams for AI-driven business transformation. Rapid prototyping then exposed the undocumented processes, conflicting standards, duplicated effort, and poor data quality standing in the way of their dreams. In that sense, AI serves as an organizational mirror because when agents struggle, they often teach you something about the business.
But process redesign is only part of the challenge. The other major constraint is change capacity. Nicole and Hillary are both operating in environments where the technology is improving faster than their organizations can realistically absorb change. They discuss experimenting with “good enough” solutions, following the energy of early adopters, sharing successful use cases between peers, building trust with skeptics, and giving leaders closer to day-to-day project work permission to control the pace of change.
The goal is to build a trusted core of people at their firms who can experiment, interpret what is changing, and act as a kind of shock absorber for the rest of the organization. This helps the firm move fast without forcing everyone to experience the full chaos of living at the bleeding edge of the AI frontier.
The throughline is that AI itself is not the point. It is an increasingly powerful catalyst for redesigning the business around the fundamentals AEC firms have always cared about: high-quality work teams are proud of, profitably run projects, happy clients, and engaged employees.
I think you’ll love this one.
▶ Watch or Listen
Watch or listen to this episode via YouTube, Spotify, Apple Podcasts or wherever you get your podcasts.
📺 🎧 YouTube
📺 🎧 Spotify
📺 🎧 Apple Podcasts
📚 Show Notes + Resources
Books
Tharp, Twyla, with Mark Reiter. The Creative Habit: Learn It and Use It for Life. Simon & Schuster, 2003. Referenced for her "banker's boxes" system for tracking ideas she's not actively working on.
KA Connect Talks
Leonard, Dorothy. "Deep Smarts and Core Capabilities." KA Connect 2017. Referenced for the "critical knowledge transfer" cohort project.
Milton, Nick. "Lessons Learned from Lessons Learned." KA Connect. Referenced for the "lessons learned databases are where knowledge goes to die".
📃 Episode Transcript
This transcript was lightly edited for clarity.
Chris: Nicole and Hillary, welcome to the Smarter by Design podcast.
Nicole: Thanks for having us.
Hillary: Yeah, thank you so much for having us here, Chris. I'm especially excited to be here alongside Nicole.
Chris: Yes, you are two of my favorite people. I'm glad that we have connected you and now we're doing this together. It should be a lot of fun. Maybe both of you could start by telling us who you are, about your firm, and then maybe most importantly, how you ended up being the person in your organization who is not only leading AI and business transformation work, but personally building AI agents and workflows.
Nicole: Great. Wanna start, Hillary?
Hillary: Absolutely. I'm Hillary Thompson. I am a principal at MBH Architects. We are about 275 people spread across six offices in North America, India, and, most recently, the UK. We do a variety of different work types, anything from multifamily, retail, labs, and much, much more. My seat covers brand communications, knowledge, and data strategy currently.
I'm not coming at this from a technology background. I'm coming at this from the perspective of watching really smart, really talented people face a great deal of friction in their everyday work as they do these really complex, amazing projects. My work in brand communications has led me into every corner of how the firm is operating today.
And the longer that I've looked around, the more obsessed I've become with answering that specific question of why are all these talented people reinventing work that we've maybe already done, or hunting for information that already exists? So AI has turned out to be the next natural step in that same question, watching how knowledge and work move through our organization.
Chris: I love that. How about you, Nicole?
Nicole: Great. I'm Nicole Chavas. I am president and one of the co-founders of Greenprint Partners. We are much smaller than Hillary's firm. We are 30 people, 12 years old. And so we grew from 10 to 30 in the last few years, which for us is a huge amount. We're a sustainability-focused urban planning, civil engineering, and landscape architecture firm based in Chicago.
Most of our work is public work, so most of our clients are municipalities, park districts, school districts, counties, that sort of thing, which really shapes the type of work that we do, the clients that we work with, and the nature of our operations. Government brings with it its own complexities, as does having three different practices, which are all relatively similarly sized, all working together in a very interdisciplinary way to achieve our goals.
And that brings a lot of complexity with it that I've had to navigate. As the co-founder and effectively the head of operations, I've been building from day one. I am not an operations person by background, and I started Greenprint relatively new to the AE industry, so I was always trying to figure things out as I went.
I built all of our systems, our hiring practices, our proposal development process. Every process that Greenprint does, I had to build at some point, steward, and hopefully hand off at some point. And so AI is just the next thing for me to tackle. Every year there's something new for me to tackle—knowledge management, learning and development, and now AI—that I have to figure out how to bring our firm along with.
But the benefit of being a small firm and being that sole decision-maker is when I see something cool that I think can help our company, I get to be the one to say, "We're gonna do it. Let's do it." So that has me really excited about all the things that are at our disposal right now.
Chris: Could you tell us, Nicole—there've been some funny jokes over the last year and a half I've been getting to know you, like you're a one-woman human knowledge management system, or at one point it was Nicole LMS. You were these systems before you had these systems.
And can you just talk about the change from being the human knowledge and learning management system to actually building systems for the company?
Nicole: Yeah. So when we started out, we were three, and then five, and then eight, and when I was building our systems and helping design the firm, I built my own knowledge throughout all of that. So even when I had systems that I was asking people to use, I was still the most knowledgeable person about how those worked and what they meant for the company.
And so over time, I still had the biggest picture, the deepest knowledge about everything the firm did, and there was nothing easier than asking Nicole the question to get the information you needed. And so even as I started doing my best to document things, to try to build training, to try to pass things off, there was still a lot of friction for folks to try to figure out how to get those.
Okay, which Google Drive folder was that in? Which training said what? It was still the fastest way to get information was to come to Nicole, the human knowledge management system. And that worked for a while because I had the answers, they were quick, I'm very accessible, and so I was able to keep doing that even as I kept trying to document.
But we just hit a point when we'd gotten to, like, the 20-person staff. We were hiring new people who were starting from scratch, and that was just not gonna be sustainable anymore. I really had to find a way to truly move that knowledge into something more centralized that people could access that wasn't so reliant on me.
Chris: Yeah, that makes sense. And Hillary, I know that your intranet is called Archie, and there were iterations of Archie before you started using Synthesis, but how did you get sucked into it? Was there a moment where you're like, "Oh, I personally can help elevate the way that our company's thinking about knowledge and learning?"
Hillary: Oh, absolutely. In my day-to-day work, as I mentioned, I do a lot of digging into every corner of the firm, and I realize more and more how undocumented a lot of our practices and processes are, how siloed they were across studios, and how lots of different teams were doing things lots of really different, interesting ways, and how much value there was in bringing that all together.
And the latest generation of Synthesis is absolutely supercharging that now, with the AI capabilities built in.
Chris: Yeah. But maybe we'll start there. Firms getting started with AI in general, but more specifically agents—their eyes get super big and they're like, "Oh my gosh, I could do everything. I can change all the ways that we're working." But then they hit a bit of a downshift, like, "Oh, but where do I begin? Do I have the knowledge I need in order to be able to build these agents?"
And I'm curious, as you've been doing your initial work—you're both in the private beta for the Synthesis knowledge agents—how did you start figuring out what problems or what opportunities you were gonna go after in terms of building agents?
Hillary: I think the most obvious answer is probably frequency times pain, right? If you're looking at a problem, something that happens 50 times a month and wastes 10 minutes every time is gonna beat something that's mild and happens twice a year. But for me, there's also a different layer worth considering, to really understand what the possibilities were and how they fit into our workflows.
For me, it was really important to start with something that I knew super well. I've personally touched thousands of proposals in my time at the firm. So it was one of those things where I knew the pain points, I knew the data, I knew where the information lived, and I had a team that I knew would really adopt it and iterate on it and come up with new ideas about how it could be better.
So that was a really natural place for me to start. I knew I could socialize it, and also important, I knew I could analyze the results myself, which really helped.
Chris: Trying to streamline somebody else's work that you don't know as well.
Hillary: Yeah, exactly. And quickly it became a matter of, is it faster and better for me to do this, or is it faster and better for me to build an agent to maybe do this and have it be a repeatable piece of infrastructure for us?
Chris: What was the first one you built where you're like, "Oh, I see how this is gonna change the way I'm working"?
Hillary: I think three or four years ago, when we all started using the chatbots, it was like, well, anyone can do this. So what edge is it giving me in my work specifically, if everyone across every firm is doing this? And it was the first time that I built something that answered questions using our firm's knowledge, rather than the public internet, that I was like, okay, this is not just a productivity tool or something that's gonna give me a one-time answer.
This is infrastructure for our team. It's gonna change the way that we actually work. It's chatbot as helper versus agent as infrastructure.
Chris: So maybe another way of saying what you said is the differentiator is gonna come because your firm is differentiated in your knowledge and your approach, and the technology's neutral. It's what you know and think and how you talk about the world that's different.
Hillary: Precisely.
Chris: Okay. What about you, Nicole? Did you have a light bulb moment for you?
Nicole: Well, I went through a journey. 'Cause I'm a builder, and I get so excited, and I love to learn. So these agents come out and I'm diving right in, and I wanna solve the biggest problems that we have, as opposed to exactly what you said—where are the places where pain times frequency could actually make a difference now, in a simple way that people can understand.
So I went too far, and I was like, what are the biggest problems? Where are there huge gaps in what we do that are so hard we've never tackled them, that maybe this agent can solve for us? So just as an example, in our proposal development process, one of the challenges that we have is that we tend to jump right in. It's like, okay, there's an RFP out, let's jump in and start writing the proposal, and we don't take enough time up front to do the real strategy, which takes a lot of knowledge, internal and external. What is our client intel like? What do we know about this client? What research can we do—by reviewing their meeting minutes and their board meetings—and pull it all together to say, "Here's what their needs are. Here's what Greenprint has to offer. Here's our win themes." And we tend to just be like, "Okay, we gotta start writing this proposal to get it out." And so it's, oh God, I'm gonna solve that with this magical agent. It's gonna surface all of that, and it's gonna create a strategy brief that's gonna bring us all together...
And what I realized was that was too much to try to do with one agent, even though it is addressing a big problem. What it really taught me was that I was trying to build a new process, create a new standard, and agentize it all at the same time, 'cause I wanted to solve the biggest problem first, and that biggest problem was just a blue ocean of challenges that needed to be overcome.
And so that's when I was like, okay, stop. This was good for me, I learned a ton, but I needed to step back and think more like the way Hillary is thinking about it—truly, where is there something that is causing a lot of pain but could be more simply addressed.
So, for example, an agent I'm working with our marketing team on is one that can simply customize our project summaries for proposals. Talking to my marketing coordinator, she does this every week: she takes our boilerplate summaries, there's an RFP, and she's the one who tries to figure out how to customize them so that we're speaking to what the client's needs are by showcasing our project work.
That's simple. It's something she's doing. It's annoying, and an agent can absolutely make that a streamlined process. So I'm learning about where to focus my energy on the agents that can get the most bang for the buck, really clear the way to better outcomes for everybody, and have people feel like they understand how to use them and they're getting the benefit from it.
And I'll tackle those bigger problems first, before I start throwing agents at them.
Chris: So there's kind of like frequency times pain, maybe divided by achievability.
Nicole: Right. Yeah.
Chris: Like that—there's this, to get to the right score. On the achievability side, though, Nicole, just to follow up on that: did you feel like that was more the achievability of what the technology can do, or the achievability of people's ability to change their processes that fast? Where was the problem?
Nicole: Yeah, no, the technology can do it, and I still see this future strategy brief creator agent, but the steps that needed to happen before that were, one, defining the standards for what it looks like to do that, and then two, actually engaging our staff—the staff who work on proposals—around collaboratively agreeing on those standards. And then creating the processes for how that fits into our proposal workflow.
And so once I've done that work, which is really change management work, which has nothing to do with AI agents—although I am doing it knowing where I want to go is that an agent can help—that all needs to be done first. Then, when we roll out an agent that can do this for us, everyone's gonna be like, oh my God, yes, this makes so much sense.
So it really needs to be staged out, and that takes longer. But the outcomes I just know are gonna be better.
Hillary: I think you have to add one more into that equation, which is repeatability.
Chris: Interesting.
Hillary: Because some of what you may perceive as the more low-hanging use cases—things that are easier for us to immediately solve—those are often the things that you can repeat across many, many different types of work. And you're also building trust, because you have something that people are telling stories about, that is working well. You have people saying, "Oh, this is how we're using this, and this is how it's changing my daily workflow." And so you get the start of that kind of internal marketing piece happening.
Chris: Do you have an example in mind, as you were talking about the repeatability?
Hillary: Yeah. So we have really grappled with the idea of a detail history library forever, and there's a big debate about what exactly we should have and what we shouldn't have, as far as people copying things. So one of the things that the marketing department has done now is create a precedent library for marketing work, which allows the team to explore the last 10 years of graphics that they've created, so that it's not a matter of, "Hey, do you remember that one thing that we did for that one thing back then?"
It's asking an agent that will pull up the five different examples that we've had throughout history, as context. And so that's something that's now super repeatable—to the detail example of, if you wanted to have a library of these documents for context and precedent, that's something that we could really easily create in the same model.
Chris: So, establishing some of these patterns of combining knowledge and process in an agent together.
Hillary: Yeah.
Chris: Cool.
Hillary: And knowing when to use the raw and when to use the curated.
Chris: One of the things, Nicole, coming back to your example—actually, I think this is gonna bounce back and forth—the taking on something that's too big, or at least too big for now. I think you said, I'm gonna take on the proposal process. But I think what was implicit in what you were saying is there's actually no one proposal process. It's actually dozens of small processes that are stitched together to go after a proposal. And it sounds like teasing apart the sub-processes and trying to improve those will then add up to a better proposal process than trying to just build the Rube Goldberg mousetrap machine that does everything in one shot, that'll just fix everything.
Nicole: Mm-hmm. Yeah, I think that's exactly right. I think there are some little small processes at the firm that could be really easy to tackle, but then there are these big ones that are multiple processes—and for us, I think that's project management and proposal development, the two things that really drive our business: winning work and managing projects well. Those are really a whole spectrum of processes.
So I really had to sit back and break down the entire length of the process into sections, and the handoffs between them, to really think about where we're doing well now, where I should just say that's a standard—we have processes in place, but did we ever really say they're the standard of good, or did we just say this is our process? And that's kind of what it—
Chris: This is what we're—
Nicole: Process. Yeah, exactly—it's what we're doing, but I don't think we ever had that question of, is this the best, that we wanna enshrine as the way, versus this is our process for now 'cause we have to do proposals. So really, it's funny, I came to this being like, "I'm gonna make all these agents," and I actually haven't built a whole lot of agents, 'cause I actually had to step back and map out our processes and help me really understand those, and make sure that I had shared understanding with our team before getting to that agent part. It's really helped me slow down a little bit and try to be more effective in how I plan to build, so when I do build, it's gonna be that much more valuable, and I can bring my team along better in the process.
Chris: How does your team feel about... I'm trying to think of how to phrase this question. I've been the person that's like, "I see the future, I'm gonna go call out all the processes, then we're gonna fix all the processes," blah, blah, blah. And it's like sometimes people are like, "Yeah, but I don't want that process changed right now. I know it, I'm doing the thing." Can you talk about that human piece? 'Cause sometimes those are processes that you're not necessarily doing—it's somebody else's work—and you're redefining what good looks like for their work.
Nicole: Mm-hmm. Yeah, that's been a learning lesson for me too, because I think for me, as the chief operating officer and a builder, I love to learn, I love to build, but most of all I want to make other people's work easier. And so sometimes I jump into, "Well, I'm gonna fix these processes for you 'cause I wanna make... And I'm not gonna bother you with it. I just wanna fix it and then have it help you, and I don't wanna bother you 'cause you're so busy." I'm doing it 'cause I wanna save people time, but that's not actually helpful. I need to really understand where their time's being spent, where the real hiccups are, and then map that onto my understanding of where I'm seeing the gaps in the quality of our proposals or in project management—places where it's not really delivering the results we want it to.
So I do have this full vantage point of where things aren't working, and ideas about how to address those, and best practices from the industry that I'm using to help think about how to reshape things. But then I need to come back. I need to take that and really sit with people and understand their day-to-day and where their challenges are that are maybe inhibiting them from being able to get to that next level.
How can we clear that stuff out first? Then they can start thinking about these bigger-picture ways to improve processes, and I can help them connect it to all of that.
Chris: Hillary, have you had any similar... You're both people I consider tip-of-the-spear people, right? But then you need to turn around and bring people along. Did you have any similar challenges, in terms of getting far out ahead of your team?
Hillary: I've been working with a really great group of leaders here who are highly motivated in this area and really into experimentation. I think in general, bringing humans along in change management can be so arduous, especially across the length of architectural projects. But I feel like we have this moment right now where AI is still a little bit magic to people, and people are really excited about it.
Chris: Not work yet, it's just magic still. Yeah, right.
Hillary: But it certainly has exposed a lot of gaps in processes, and a lot of places where a human has been quietly patching over the missing pieces, the missing standards, with good judgment. But AI has really exposed—if I have instructions to follow this implicitly—what's missing.
Chris: Yes. Meaning like you could have a checklist with seven steps, but those are just the high-level steps, and humans are filling in 7A, 7B, 7C, all the other—
Hillary: Exactly. But you also have this rare moment—we've all written instructions, we've all written emails, right? You're like, "Maybe 50% of people are gonna read this," and 10% are gonna make it to the end of this email or instructions or follow it to the letter. But you also have this motivation moment where you're writing instructions for something that is gonna follow those instructions to the letter. So it's also a different moment in that way, a different motivation in that way.
Chris: I had never put that idea together. People may or may not read the instructions you give them on how to do the project, but the AI will, and that's both good and also it's gonna read everything that you said or didn't say.
Hillary: Yeah. The instructions have to be of a different quality, and your processes definitely have to patch in those places where they've missed stuff previously.
Chris: I'm glad we're talking about this. I feel like inside firms there's a spectrum of the AI-is-magic spectrum. I mean, I think even the most knowledgeable person that works in a foundation lab still says, "We don't know exactly how this works, but it's amazing," right? But there's a continuum of, "No, it knows everything, it can do everything, it's all-knowing, it's AI." And the other end, they're like, "Eh, it's probabilistic, and we know how it works."
And I'm curious, as you try to communicate about the need to change processes and capture knowledge and get more intentional about documenting what good looks like—have you found people who don't buy that? Who are like, "But AI should just be able to figure it out. Why are we spending all this time writing this down? We have this super intelligence in our company now."
Hillary: Oh, I think almost the opposite. There's still a skeptic group that's like, "AI is not gonna figure this out. AI is gonna get this wrong." And those people are right, right? AI is not getting everything right today. Those people absolutely belong in this conversation and need to keep telling the firm that the human being must remain in the loop.
You are ultimately responsible for these drawings and your work product, not this system. And I think the balancing of those two things is really good. I recently turned my focus to—I guess you'd call it—putting a fence up. Like, when you have a dog and you don't have a fence, you have to walk the dog on a leash.
I think when you put better guardrails in place, when you put a fence up, you can let the dog roam free. I just recently turned my focus from, yeah, we have this AI policy that we've had for years and years, to, let's write something a lot more practical that tells people what they can put where, so that they can go wild on these tools and start using things.
And that also balances the worries of the skeptics a little bit with the people who are highly motivated to do something but maybe need to be cautious about putting the right thing in the right place.
Chris: So, in my understanding, Hillary—it sounds like you have more of the skeptics than the over-optimists, in the population that you're talking to about this change with, or do you just have both?
Hillary: I think the skeptics are just louder in a lot of cases. The people that are really into it are quietly working away, and "look at this cool thing that I did" occasionally. The skeptics are maybe like, "Be careful, be careful." They're the ones a little louder in the room. I don't think that we have more—I think it's probably less skeptics than we do people who are excited. But I think it's the people who aren't saying anything that you really have to worry about bringing along.
Chris: How about you, Nicole? How are you thinking about that, in terms of the different cohorts or communities of archetypes of people, as you're thinking about the change management that goes along with the building?
Nicole: Yeah, it's interesting. We started engaging the staff around this two years ago now—we wrote our first responsible AI policy just to lay down some ground rules. It was such a hot topic. You could kind of start seeing... And this was, I think AI's gotten so much higher quality since, but this was kind of the earlier ChatGPT days, when everything felt even more generic, and people could really see it in Slack—it'd be like, "Oh, that's so obviously a Chat..." You could see people were testing things, and people were feeling uneasy about it. There were a lot of questions.
So we have an annual staff retreat, and we dedicated a day to AI, so we could surface from everybody where they were excited, where they were concerned, and get a sense of where people were at. We could see the people who were super jazzed, but I'd agree, it was more skepticism. But I think that went a long way in building trust with the team, us saying, "Look, this is new. Everything's changing. We're exploring it. We will always stay committed to our values. We have created a responsible AI policy, human in the loop. We stand behind our product. We will disclose," those sorts of things.
And then, as the tools got better and we got better at using them—particularly around what Hillary said, using it to harness our own knowledge as opposed to just create generic content—people started to see that. Our reporting got so much better, because it got more finely honed on what actually matters. Our ability to better tell our story in a way that's compelling to clients, because we're able to use AI to help us make connections.
I think people could start to see those benefits and realize, oh, this could actually be useful. So we had a bit of a natural experiment recently to test where people are at with that: we use Claude, and I decided to get Claude Teams for anyone who wanted it. It's opt-in, and I'd say about 35% to 40% of our team opted in to that next level. Everyone can still use Claude Free, subject to our policy. But for the people who are ready to go in and really test out the next level, it was just available. And it's really interesting to see which teams were the ones to jump on it and which were not.
And I'll say the folks in the most technical engineering roles were the least likely to jump at it, and the ones in the more creative roles, or the more engagement roles, or the broader client-facing roles, were the ones to jump on it—because they could see how it could help them with contract management, how it could help them build a plant library, how it could help them synthesize research more effectively for planning efforts.
And that's helped me see, outside of my own little box of how I use it, how our teams are using it. So I beg them to show me the things that they're building, so that I can learn from them as well. It's very much—we're open, experimentation is welcome. We have a tool that's enterprise-level now, but no one's being forced to do anything.
Chris: That's super helpful. So, going back to frequency times pain times repeatability divided by achievability—this is gonna be our running bit—there's a question I know you talked about earlier: the technology can do it, but I'm thinking about that group that opted out, Nicole, and that's not unusual. I've heard this across other engineering orgs too, where maybe those people tend to be somewhat skeptical just as a breed, and/or the technology in that specific instance isn't as far along as it maybe is in some of the other ones, and maybe it's those two things combined. You're nodding, for those who are listening.
And I'm wondering, what's the mindset? Is it, you know what, it's 2026, we're at the beginning of this thing, who cares, we're gonna move on and make improvements elsewhere, and maybe when we come back people will be more excited about jumping on? Maybe the tech will be better? I'm curious how you're thinking about the uneven adoption of AI within the company.
Nicole: Yeah, I think that's it. I think it's about how it—beyond just basic chatbot stuff. I know everyone's using chatbots for something or other in simple ways, honestly just as an advanced search. So I know everyone's doing that as a baseline, but for many people, the free version works just fine. They don't need to go beyond that.
I think it's a mix of them not seeing, in their day-to-day, how something more sophisticated than the basic search really helps them do their job better—relative to the people who do see, "Oh my gosh, thanks to this, I can synthesize this information very quickly in service of moving a project forward."
So I think it's that. I do also think, and this is a challenge for me too, until you see someone do something cool, you don't even know it's a possibility. And it's also about time and capacity. One of our landscape architects—she's one of our most gung-ho, and she has better skills than I do—she's building out this plant library, thousands of plants that we're assessing to choose for different projects, and she's got them all categorized by the sunlight they need, when they bloom, all these different facts, type of soil, and it's this really interactive visual library that she built entirely in Claude.
I'm sure the engineers could use something like that, but they're not thinking about it. They don't know it exists. They aren't thinking about how to do it themselves. So figuring out how to get more skill-sharing across the team, where they can show each other these things and help each other out—that's something I'd like to do more of: create more spaces for our team members to come together and share their creativity and see where it can add value, versus me saying, "Here's a cool thing. Don't you see? Don't you want to use the thing?"
Chris: Mm-hmm.
Nicole: So that's where I hope to get to, as this becomes—well, nothing's ever gonna become—we always joke that everything's always in beta, so we'll be in beta forever, but as it becomes a little less beta-y and a little more day-to-day,
Chris: Whatever comes.
Nicole: Yeah, exactly.
Chris: It's been a minute, Hillary. Let me step back. I agree that people need to see that something's possible, and I guess the more it's their work, the example they see, and the less translation they have to do, the better. But in your language, I think it was, I'd like to do more of this, but it also doesn't sound super urgent—like you don't need to be in a rush, and if they're not ready yet, that's fine, we're still learning. Is that true for you too, Hillary? Are you taking your time with some of this transformation work?
Hillary: Yes and no. I have been taking my time, but I typically lean towards more cautious and measured in my approach to things. And with this, if I could give my earlier self, months ago, some advice, it might be: start faster, start sooner, move faster. And just know that when you hit a roadblock, it's probably gonna be solved within weeks, it's gonna be different. The pace of things is just moving so quickly now that I'm trying to be less of myself with this, more in experimentation mode. I think also the build time is so reduced now that there's not so much investment in trying things. Getting a proof of concept out doesn't require you to have everything perfect in the background—just knowing that it could work if it's something that's gonna hold value to your work.
Nicole: I think that's a really good point too, just as our teams get more sophisticated, 'cause I've been feeling that way too. I don't wanna release anything unless it's the perfect agent that I've tested. But when I sit down with my marketing coordinator, she gets it, and so it's, oh, if it's 80% good, but then she has the judgment to take it to the finish line—well, let's solve 80% of her job. I don't need to make it perfect for her, and we can iterate on it together.
So how can I be okay with rolling things out that are good enough, and trust my team, or give them the guidance that it's not perfect, you should look out for these things, but here's how you can take it to that next level, or we can iterate on it together, and enable more of that experimentation and testing collaboratively.
Hillary: Yeah, different approaches for different people though, 'cause you also have people where trust is a ledger, and you really need their first experience with agentic AI to be positive and valuable. And then you have the people who understand it, have the vision for it, and you can show them 80% and they will take it to 150.
Chris: This feels like a bit of a tangent, but I've been hearing, in some of our clients talking to their leaders, a theme over the last six or 12 months that I just wanna test with the two of you—and it may be that you're not seeing this at all, and then we can move on, or even cut it—that the pace of change just keeps accelerating. Hillary, you shared that even for you personally, it's like, I need to let go of some of my ways of working in order to be in this moment and take advantage of it. But there's some pushback—some of the firm leaders I've heard back from, some of their staff saying, "There's too much change. Things are going too fast. We need things to slow down, no more change." And they're like, "I don't know what to tell you—this is the least disruptive it's ever gonna be." Again, we don't know that, it could very well just keep continuing.
And so people's willingness to pick up their pace, or become a little more good-enough versus perfect—is that it? Are those conversations happening in either of your firms, around who we need to become as people in this new era?
Nicole: That is 100% a challenge we're having, and we talk about it every day—how much change there is, how challenging it is, and how our team is feeling the pressure of all that change. But we can't control everything. I think we can do a better job. There are some things we could let go. I think we need to prioritize where the change truly needs to be. But we're all in the pressures of change. The world is so different, and I think it's gonna be a hiring criterion in the future for staff to be flexible and able to manage change.
I think we've lucked out in that that's always been something we've looked for, because the nature of our work—working in sustainability, across these different practices—things change a lot. Our clients are changing a lot. Working in government in this current environment, things are changing a lot. So if you can't be flexible with how the winds change, it's hard to work in this field. We've always kind of looked for that, but now it's like times 100. So even our most flexible people are feeling this whiplash. I don't have the answer, but it's top of mind, and I'd love to get the community helping figure out what it looks like to address this in a sustainable way and bring teams along.
Hillary: I think here at MBH, it's maybe more about segmentation and platform fatigue—having to know which platform to go to to get information, or there's this new platform where you have to enter all of your milestone dates, or something else where you have to do your resource planning. There are just so many different spaces where you're now entering data. I think the wonderful thing about what AI will do for us is it allows us to actually unite all of those sources in the future, and use all of the information that's already in disparate places. So I think connectivity will help with that.
But also, the wild thing about AI right now is that when we're talking about the future, we're talking about three months from now, not five years from now. It's kind of, it's here, grab on, because we have to go at this speed now.
Nicole: I was gonna say, it's really hard for me, 'cause when I see a better solution, I wanna put in the better solution. We have this conversation where I'm like, I think we should do it this way—yes, we were doing it this way, we were testing that out, that's not working, I think we should do it this way. Or, I shouldn't say it's just me and my whims, but we learned something from trying it this way that's telling us we should actually do it this way. And then it's like, yes, that would be better, but is it gonna be too much change to then tell people we're changing it to that?
So there's actually this weird thing of, we know that will actually be better because we learned so quickly by doing, but then we're asking people to change again. This is silly stuff, but—meeting cadences. We keep trying to figure out the right meeting cadence to get the right people together in the room to share the knowledge when it's needed. And we keep trying to figure out the best way—when should we engage on Slack versus in person? We keep changing things around, and we learn every time better ways to engage with our team, but that shapes the way we think about doing it differently. But then people just have no idea what's going on.
Chris: So it's like, if this change could make us 30% more efficient, but it's gonna cost us 25% to do the change management, is it really—
Nicole: Yeah. Or do we lose some trust, like, "Oh, they're changing it again," even though... Yeah, exactly. It's this calculation that we have to do.
Chris: So frequency times pain times repeatability times—
Nicole: Loss of trust.
Chris: —loss of trust might. Yeah, carry the two. All right.
Hillary, you did something in the nexus of things we're talking about—doing a design charrette with your leadership team around starting to help them envision how agents could work and what that would do for them. Can you talk about that charrette a bit?
Hillary: Yeah. So we had that full spectrum of AI knowledge in the room, from skeptic to people using it at a pretty impressive level. Everybody brainstormed individually what agent they would want as part of their work—I gave them five minutes to do it, so they could just come up with as many as they could. Then I put them into smaller groups to discuss those things together and come up with what they thought would be the most valuable for the firm.
And reading through people's individual answers, it felt a little bit like people putting their work dreams on paper without really any expectation that they would ever become a reality in many cases. It surfaced a lot of really real problems that people are facing in their daily work, a lot of really good trends for knowing where to build. So that was week one. Then, between week one and week two, I built three of the agents that different groups had come together and agreed on. And that was what really changed the conversation. It became not this abstract thing, or some distant promise, or faraway work dream—that was when people were like, "Whoa, that quick?
Chris: Mm-hmm.
Hillary: That came into reality that quickly?" So the tone really flipped, and then people in the room, instead of debating the value of AI in an abstract way, started asking really specific questions about how they could apply it to their work. It was like they suddenly went from spectator to co-author in that moment. And if there's one thing I could stress to anybody doing this, it's bringing those real working examples as fast as you can.
Chris: To show the prototype of the work dream.
Hillary: Yeah, absolutely. It doesn't have to be fully baked. That said, I do think having one thing that's pretty well along—pretty fully baked—at least, is good to show them, so that you get that trust going.
Chris: How much context did you need to give them to be able to have those conversations about what an agent is and how it works? Did you give them any, or did you say, "There's this technology that can make your work dreams come true. Go think about what your work dreams are"? How did you...
Hillary: I guess maybe a 10- to 15-minute spiel, introducing the difference between an AI chatbot, agentic AI, Synthesis AI search, Synthesis knowledge agents. Gave them a brief overview, and then also a kind of check chart—this does this, this doesn't do that—so that it was easy to follow along. And then I intentionally left it kind of open, 'cause I wanted the creativity and the possibilities to come.
Chris: You didn't try to give them, "These are known limitations of the tools today." You just said, "Go, go do it. Write..."
Hillary: Yeah. 'Cause that's also how we're gonna figure out what the future is—without the limitations in place. What would you do?
Chris: When you went away and prototyped the workstream pilots—I'm guessing that, back to Nicole's point from the very beginning, you didn't necessarily have all the standard, well-documented processes of how things work, or all the knowledge or documentation you needed to build them. So did you fake some of that in order to show them the prototype of how it would work? How did you navigate that?
Hillary: So, let's see. In one case, I approached one of the teams and said, "I've been hearing that you have this document where you've gone through this set and written down everything that you look for when you're checking this particular set of drawings. Could I have that document? And can I also have some documents that you've checked previously?" In some cases I just got a very small sample set. In another case I already had a good-sized sample set, and it just really worked out that that was what people wanted for that particular topic.
Because I work across the entire organization, I have a lot of knowledge of where different things live and who is doing what, so it wasn't terribly difficult for me to source the things that I didn't have. And then, in the case of that last agent, I made sure it was something I had already been working on for a considerable amount of time and had kind of perfected, and I could show them many, many different conversations with the agent, to show them this isn't smoke and mirrors—I'm using this every day, and this is how it's working.
Chris: Did showing them those prototypes build any buy-in for the change management you'd have to do around documenting their processes more thoroughly, to build some of those other dreams? Did that connect the dots for people—like, "Oh, I want that, but in order to get that we have to do the knowledge and process piece"?
Hillary: At the end of the second session, someone in the room actually said, "It sounds like we need a lot more documentation to make these work." I think I stood up at the end and was like, "There it is." So yeah, building the agents became an organizational mirror. It exposed the undocumented processes and undocumented knowledge.
The conflicting standards—that's actually a fun little agent I've built now, to compare standards across the firm and show everyone how different they are. Duplicated work, poor data quality. And helping people understand that if an agent is struggling, it's teaching us something about our business, not about the AI, probably.
Chris: Hmm.
Nicole: That’s so profound. I feel like the idea that we had to go to AI to learn about our own gaps and our own way of doing business. That idea of AI as the organizational mirror really resonates with me. Because to try to tell the robot what to do, you have to be so precise, and then when you realize you don't actually have the right words to tell it what to do, and you're projecting your own things onto it, it really does highlight all those failures, and if you keep going, the agent itself will fail.
So I just did not expect that on this journey. I thought I was just gonna be building all these cool agents that were gonna solve everyone's problems, and instead it's actually just completely turned around and made us revisit the business and the way it's set up and designed, from back to the beginning, to take us to that future.
Hillary: Yeah, and it's funny, 'cause even in the things that we already had—like I mentioned, the set that had everything outlined for someone checking the set—it was like, are humans using this? Are we using this? But it goes back to the idea that AI is gonna follow the directions. It's gonna follow them meticulously,
Nicole: That's another thing that I've noticed too—we have had some standards, they're very important standards, and they may be completely unintelligible to anybody but the person who wrote the standard. One simple example of that is our writing guide, which our marketing team, the way we write needs to look exactly like this, AP style, it has this voice, and they spend too much of their time going back and fixing all of that. But they had, in the comments, our detailed writing guide, and now anybody can do it. That was one of the simplest agents I made: review this against our writing guide and fix it. And now marketing will never have to do that again.
So it's like we had the standard, but it was unusable for most people, and now it's completely usable, and no one actually needs to know the details of AP style to make sure we're aligned with the standard.
Chris: That's really interesting. So, better writing for the AI means better writing for the humans—but not just humans in general, the humans who aren't the subject matter expert who owns the thing. So you need to make it more democratically accessible.
I'm curious about this process. One of the things we've talked a lot about in the community is not only what good looks like being important for AI, but also for doing learning and development. If you don't understand how we do project management here, how could you teach it? The way you could is you'd teach the way Hillary does project management, or the way Nicole does project management, but not the way your firm does project management.
I'm curious if that convergence has hit your firms yet, where you're thinking about learning and development alongside AI, like there's a Venn diagram that connects all of it.
Hillary: I love a Venn diagram. Yeah, one of the things I think AI is going to do most for us—we're not there yet by any means—is connecting the just-in-time learning and the hour-long training and learning. Actually allowing people... Yes, they may have completed the training, they may have taken the test, gotten an A+, but three months from now, when they need that information, are they gonna remember it, or surface it, or remember where it is?
And now that content doesn't have to be separate from our L&D content. AI has united them and allowed people to find those snippets of information in their day-to-day real work, that they can connect back to the learning and development, and that's gonna be pretty amazing when it comes into reality here.
Nicole: Yeah, I think for us—project management, obviously, is really important, and the ability to think about higher-quality project management processes and higher-quality learning and development, someone's gotta be accountable for that. And as we've heard, I do too much, and I can't—I'm an inch deep, a mile wide. So I think at the end of the day, to get to that, it can't just be me. One of the things we did last year was actually tap one of our practice leads to take on a hat where she's the vice president of practice operations. Over our three practices, she's responsible for project quality, setting our project management standards, and making sure everyone's adhering to them across the practices, 'cause that's exactly what's happening—each practice had their own way of doing things. This is the challenge of three co-equal practices doing different things, but there are things we should all be doing.
So that's step one—having someone responsible for that, who felt accountability for it and committed to making a change. Then there were two sets of what good looks like that had to happen to get to better project management processes that people are trained on and actually adhering to. First was having her revisit our standards, our processes, and refresh them—she did that by interviewing all the staff, collecting feedback, and then doing that refresh. The second was a training, which also has a what-does-good-look-like for how training is designed and delivered, which was also new to us.
Chris: Right. Very meta.
Nicole: The idea of what good looks like for a process like that, I feel really confident in, but what good training looks like was very new to me. It honestly wasn't until joining this community—having the idea that I have this great LMS at my disposal, and learning about adult learning principles, and hearing how other people were structuring trainings in terms of length and course content—that made us completely rethink what a good training is. Historically, our trainings have been very much, slide decks, here's the process, screenshot, do the thing, okay, now you've been trained.
So we had to do the same thing. What does a good training look like? It's based on these adult learning principles. Here are our ground rules. A course should be this long. It should have this mix of optional and required content. There should be a fun homework item in the middle. It should have a quiz. There should be expert interviews in it that keep people engaged. It shouldn't be more than an hour. Those sorts of things are our new standards, and she, along with some team members, was able to build on that—the output is so good, and I had nothing to do with it. Maybe that's why it's so good.
So it was these two-step what-is-a-good process to get to these higher-quality trainings that we're starting to see leading to higher-quality project management effort. It's really exciting to see how that all plays out. It's all related—the processes alone aren't enough, the training's not enough. It has to be a fully thought-out approach to improving the quality of our work and getting to better budgets, better schedules, happier clients.
Chris: I mean, the phrase that's running through my mind as you're telling this story is—just even for your example, Nicole, at Greenprint—redesigning proposal process, redesigning project management, redesigning the way that we learn. In a couple of years, will there be anything left from... Do you know what I mean? Are we redesigning our businesses from the ground up during this transformation?
Nicole: Uh-huh. Yeah, I think so.
Chris: And it's interesting that—I'm curious on the project management piece, 'cause we talked about the proposal one, trying to redesign the sub-processes versus the whole thing. And when you were focused on project management, which is also, like, proposals, that's a big bucket of things.
Nicole: Did you, from a training perspective, go after it on specific subsets of project management first? Are you tackling sub-processes first? That's exactly right. When you actually map out all the phases of project management, there's contract scoping and contracting, there's the launch process, there's the monthly project management maintenance and processes, and then there's closeout. We decided to focus on monthly project management, 'cause that's the bulk of the effort—that's what people are spending most of their time on. And that's where we saw a lot of opportunity for improvement: a way to get higher-quality data, better elevate problem projects that need to be addressed, and get to real consistency in the client experience.
And then from there, I think it'll be project launch and contracting, to an extent—although contracting's good enough now, I'm not as worried about that as the monthly client experience and how well we know our budgets. So it was basically looking at all the sub-processes and focusing on the one that was the highest impact, that people spend the most time on, the most repeatable—it happens every month, as opposed to every time a project launches. It just felt like the most necessary to focus on.
Chris: It looks like Hillary's about to say something, but I can't tell if she is, 'cause I have a follow-up question, but I could see you calculating, so I didn't wanna jump in.
Hillary: Now please, please ask your follow-up question.
Chris: Okay. I wanna stay with the redesign. This is like outside-in, what I think I'm hearing. On one stream, we're rethinking proposals piece by piece for an AI world. In another stream, we're rethinking project management piece by piece, but more from a how-do-we-train-people world. But is the opposite also happening? When you're thinking about the proposal piece, are you also thinking about how we're gonna train people to do proposals? And when you're doing the project management piece, is there an AI component—how do we agentify some of this project management work?
So are you thinking about learning and AI at the same time in both streams?
Nicole: Absolutely. We approached them from different locations on that spectrum, because I think we started with learning on the project management side, and then we were like, "Wait, we need to clean up these standards first." And we used AI throughout the whole way—to write clean, clear standards, SOPs, and design the training, all that sort of stuff too. So AI was used all the way, but now, because we have this refreshed set of key templates and key processes, I see the future plug where the AI agents will help.
Chris: Mm-hmm.
Nicole: And so then, with proposal development, I started with agents. I was like, "Oh yeah, agents are gonna fix all of this." And then I was like, "Oh wait, we need to address what good looks like." So now I'm nailing down what good looks like—that's gonna help refresh our processes, and now I'm gonna do the same thing I did with project management. We're gonna get those trainings, and now I've got a template that really worked for project management that I'm gonna bring to proposal development, to bring that same perspective.
So I entered both of those streams from a different part of it, but the end result is gonna be the same: clear what good looks like, clear standards, really good processes and templates, and then AI agents that can plug on to improve those processes for everybody.
Chris: And then potentially training. Yeah.
Nicole: And then more training. Yep, just constant.
Chris: Yeah, that's it. But the what-good-looks-like is the gateway to have those options.
Nicole: If that's not there, the whole thing falls apart.
Hillary: I got stuck on the word redesign there for a second.
Chris: Is that where you went, before, when I saw you? Yeah.
Hillary: Yeah, I was really thinking about the word redesign, because... the way that we're getting there is redesigned, for sure—overhauled. We were swimming there before, and now we're in an airplane. But I don't know that the destination is that different. I think that the possibilities have unlocked what we always knew to be true, who we always wanted to be, what we always wished we had time for—it's suddenly coming true.
With L&D, all of a sudden we have all of these capabilities at our disposal that make it easier for people to impart that knowledge without so much effort, and to make that super scalable across a much larger organization than it used to be. I think we always knew that we wanted to have these standards. We wanted to teach people these standards. We wanted to be the best project managers that we could possibly be. We wanted everybody to do it roughly the same way, with their own twist. We just never had the time and resources before—now, with these new possibilities unlocked, it's like boarding the plane and getting there in two hours, instead of having to walk or swim across the ocean.
Nicole: That's so interesting, 'cause you're right. At the end of the day, what we're building these on are industry best practices. We're not reinventing how project management is done. We're just actually making it work for Greenprint—having it be clearly communicated, easy to actually execute, and training people on it.
I think about some of the parts of the process that AI is now helping us with that have always been best practices. In public procurement, you've got all sorts of publicly available data that could help you better understand the client, but in the past, some human was hunting through municipal board agendas to try to find insights into where their money's going, and now AI can get that for me in three minutes. All of a sudden, this thing that is best practice, but that no one except big firms with lots of human resources could do, we can do. We're actually able to do the best practice because, with the limited resources that we have, we can just do it so much faster.
And I feel the same way about Deltek reporting and all these things—as a small firm, it would just be frustrating that we couldn't get high-quality reporting, we couldn't do high-quality research because we didn't have the resources. Now we do, and we can actually do it.
Hillary: Yeah. And maybe the piece you're unlocking really is, people can now use that time for more creativity, more innovation. Maybe that's the real redesign—all of that other stuff that we've always had to do can now go to a greater purpose.
Chris: That's so interesting, I love this, and my mind was going the opposite. I got stuck on this—the destination's not that different. I think in all the examples you gave, that's true. But then, of course, I started trying to think of the devil's advocate case. One place my mind went—which I could still get spun around into arguing the destination's not different—I'm thinking, Hillary, of the story you shared with our community around your retiring CFO, the person who reviewed non-disclosure agreements, and we can get into the details of that.
The thing that happened is you were able to rapidly do critical knowledge transfer from someone who was leaving the company, which is something that—at KA Connect 2017, we brought in an expert in critical knowledge transfer, and she did a six-month project with four of our clients, and we all shared what we were doing. Some firms did some things, but generally it was like, man, that looks hard. It's hard to scale and hard to sustain, but it was the best available technique at the time. And I look at what's going on in our community over the last few months, and I'm like, "Oh, this is critical knowledge 2.0."
We're finally gonna be able to do these things we talked about doing. It could be possible that the firm doesn't have to keep learning the same lesson over and over again as the people switch. We can actually scale some of it—not all of it, obviously, people have 30 years of experience and reps and tacit knowledge—but the amount of knowledge we could potentially hold onto as organizations seems like a step change different. Otherwise, we would've had to relearn how to do these things all over again from the ground up when somebody leaves the company. Are you feeling that? I guess, Hillary, your company is older than Nicole's. Nicole, I don't know if you've been worrying about critical knowledge transfer with people leaving.
Nicole: I guess it's been more like scaling things from experts. We have a strong demand for a robot Jim. We didn't do a charrette, but on a monthly call I asked people to put in the chat: if you could have an agent, what would it do? And multiple people—Jim is our most experienced, 30-years-of-experience engineer—multiple people said, "I want robot Jim, so I can get into his brain."
Chris: What do they want out of his brain?
Nicole: They just wanna hear all of his stories—every project risk he's addressed, how he thinks about different challenges, how he works with clients, all the things they know he has and that they get to experience on the job, but not necessarily in this really focused way. And that's on my mind too—time is always a challenge, but at some point we need a structured interview series with Jim that can help share that knowledge with the team.
Chris: Have you told Jim that people want a robot version of him?
Nicole: Oh, yeah, he laughs. He's so modest—the answer's always, "Nobody wants that." And it's just, people are crying out for a robot Jim.
Chris: How about you, Hillary? Are you thinking about robo versions of your employees?
Hillary: Oh yeah, I've also approached someone about being a robo Kevin. I think he's coming around to the idea. One of the interesting things was, after Nicole and I had the session with our fellow beta testers of knowledge agents, and heard about robo Marty for the first time, that sparked a lot of ideas. And as I was talking to Kevin about robot Kevin, it occurred to me that robot Kevin could actually help Kevin be Kevin faster, because he's not typing the answers to the questions hundreds of times over history. Perhaps these things could actually help save our experts some time also, in being themselves and answering questions throughout the firm.
But I also think the really cool thing about the capabilities now goes back to the idea that we could have 10 hours of Kevin talking—and in fact, we have recorded all of his lessons from our corporate university over the past eight years. So we actually have hours upon hours upon hours of Kevin imparting his knowledge to people throughout the firm. We don't have to now go through 200 hours of footage, come up with the exact questions, and expect people to remember everything from them. We can put them all into a library, they're transcribed, people can ask it questions, we can have agentic Kevin answer those questions. And Kevin, luckily, will still be here for a long enough period of years to make sure that robo Kevin is answering those questions correctly for people.
So I think it's super exciting. Probably a little creepy, if you're the person who's gonna be agentified.
Chris: I've gone through the experience. There's a tool that came out called Delphi, which is meant for doing this more public-facing—it's for authors and people who get tons of email, and they're like, this thing could answer email for me. I was just curious to see how it works, so I gave it all my podcasts and newsletters and a bunch of stuff I'd written, and it was really good, but I spent a lot of time with it trying to find the holes. I'm like, "I don't think I've ever talked about this." So I'm trying to trip up RoboChris. It's a really interesting experience. I've heard that Epstein Uhen—shout out to them—is using AI Marty a lot to try to break it and figure out where there are holes in its knowledge.
So yes, if the person could be the... they realize it would help them if it were to work, but I think they'll also have fun trying to trip themselves up with it. So critical knowledge transfer is getting a whole new rethink. As you were talking, Hillary, I seem to connect some dots—did Kevin also run the Lessons Learned program at MBH?
Hillary: Yes. All of our retrospectives. Yeah.
Chris: So the... we had a—actually, was it the same conference? It was right in that neighborhood, either 2017 or 2018. We had a person, Nick Milton, well known for lessons learned and his thinking on that broadly, and knowledge management, not just for AEC. One of his key slides was, lessons learned databases are where knowledge goes to die, for the exact reason you just shared, Hillary. Who's gonna think, on their project, to go look at things with exactly the right question, pull the retrospective—that was so true. Except all of a sudden I'm like, "Is it?"
You've got a database of that, and you can search it in an unstructured—to geek out for my knowledge management people—semantic way, so you don't even have to use the right term to go back and pull the lesson. Maybe that wasn't a big waste of time after all. Maybe it was just that we didn't have the right technology to make that knowledge available.
Hillary: All those shoeboxes under the bed.
Chris: All the shoeboxes all of a sudden look really good right now. Yeah. Have you been thinking about lessons? Because I know you have a lot of lessons-learned content, Hillary. Have you been thinking about whether this gives that a new lease on life at all?
Hillary: Yeah, particularly around... well, lots of different ways. I think across projects, if we can socialize it properly, people will actually start to query those to understand how we've solved problems before. But I also see that more broadly, it can help us recognize greater trends across our projects—even cultural things, like where are designers feeling pain points, where are job captains feeling pain points that we need to adjust in our processes, or technical issues that are repeats, which inform our L&D strategy for technical content. It's gonna be so much easier to identify those trends across all of that content, and also reuse the things that are reusable on a project-by-project basis, and take them out of silos.
I don't know anybody who currently gets the lessons learned outside of their team, much less studio.
Chris: Mm-hmm.
Hillary: Making them more broadly available across the firm—sky's the limit there too.
Chris: Right. And you don't—'cause the old thing with the lessons-learned database, and why it was a broken idea, is they were usually a separate thing. As a knowledge seeker, you have to think, well, is there a lesson learned on this? Is there something in our learning management system? Is there something in our project archives? And I think the world we're building towards now is, I don't have to think about where the knowledge lives—I just have to have a good question, and bring the right intelligence to me, which also makes me think the adoption of such an approach feels a lot more realistic than it's ever been.
Hillary: Yeah, I completely agree.
Chris: Where do you think, for the two of you—what's next in the future? And by the future I mean three months. What are you gonna be working on for the rest of the year?
Nicole: Yeah, continuing to—I kind of mentioned this year was the year of focusing on our sales process and our project management process, and just trying to get them to the next level, and that starts with, what does good look like? What's the process, and then what's the training? That's gonna be a continual process that we continue, as you said—we've got sub-processes, so it's creating a plan for what a reasonable rate of rollout is for attacking each of those sub-processes and rolling them out. That's probably, with the change management side of things, gonna take a while—probably another year, to be honest, 'cause really, that's what's slowing it down, the change management to actually get people trained, adopted, all that sort of stuff.
And then I think the other thing is, actually really starting to measure the results. I keep building and building and building, and I don't wanna build for the sake of building, and I don't want people to constantly feel like something's being rebuilt. So I hope that we get things to a place of good enough where people feel like they understand it, and then we start to see—are projects more profitable? Are we getting better client satisfaction? Are we increasing our win rates on our proposals? Actually looking at the data to see if the investments we're making are getting the return we hoped they would.
And then I've been so focused at the business level—business development, project management—I wanna get into the practices and start getting into, what can we do for the engineering team? What are your challenges that I could really jump into? One, it's what I know, but two, I think we have to get that company foundation right—everybody has to be doing the same thing, sales has to be bringing the work in. Once I feel like that's in a good place, then I really wanna jump into the practices and be like, what are the things in how you deliver the work that I can help use all these tools to create new standards, develop better processes, and figure out how to create the kind of training that can support you all.
Chris: Is one of the metrics you're looking to measure employee engagement or satisfaction with the way their work is being done? Is that something you're thinking about too?
Nicole: We do an annual employee engagement survey, but I don't know that it really reflects this question of speed of change or AI. I think there are new questions we need to start asking that get at the modern workplace, the modern learning organization, to better get a sense of where they're at with that. So yeah, I think that might be due for a refresh in how—what kind of questions we're asking around that.
Chris: Yeah, 'cause I was thinking specifically about your proposal process one, and Hillary can speak to this in much more depth than I can, but from what I hear, that's not always the most fun process to work in. It can be deeply frustrating spending lots of time hunting for information, and even if it didn't necessarily move the dial on the win rate, if the people who work in marketing actually enjoy their jobs more, that feels like we've done something helpful.
Nicole: Yeah. Even just so much of the marketing team's job is bugging PMs for information, and nobody's happy. Marketing just feels like a nag. The PMs feel bad that they can't be helpful, but they don't have time for this, it's not a priority, and the information they're giving them is probably just, "Yeah, that's fine," as opposed to really thinking about, what's the best information I can tell you about this project that you can use for this proposal?
So those are the kinds of things an efficient AI process, that can pull all the knowledge we have about the project before even having to bother the PM, could empower our marketing team to feel like they can get the information they need first, understand it, write it in a way that's meaningful, and just get a check from the PM. And the PM can save time and only be brought in when we really need them to bring a unique perspective.
So that's a really good point—how can it truly make people's jobs better, and have them stop having to do the stuff that doesn't make anybody happy but is a requirement of being able to do proposals or project management? How can we make things easier for them?
Chris: Quality of life.
Nicole: Yeah, quality of life.
Chris: What are you thinking about, Hillary, for the rest of 2026?
Hillary: And on that note, quality of deliverable too. I thought it was gonna take us a long time to get to that place, but a week ago, I used someone's notes from a site visit, a project description from an RFP, and a bunch of past approaches that person had written, and it provided a pretty dialed-in starting place, to say, "Here, go. Hopefully that saved you two hours of figuring out where to start."
Nicole: And the other neat thing about that is it makes the PM feel seen. "Oh, you... yes, that is me, you guys have been listening to me, and I feel seen." It's a little bit validating to feel like you're doing the work and you don't need to repeat yourself.
Hillary: Yeah. And you are our client—we want to make things as easy on you as possible. I think for me—I don't know, I guess I'm gonna go maybe three, six, 12 months here. So three months: I talked about building the fence. I didn't know that I needed a fence until recently, which is—
Chris: All I can hear is When Harry Met Sally—am I the dog in this scenario? As soon as you started talking about the—
Hillary: I have to tell you, I really love my dog. I didn't know how much I needed this until recently, which is why I keep coming back to it. But you can't have people go free unless you've provided this safe zone to go free. And particularly within my firm, we have a lot of really sensitive information and a need for guardrails, and I think we're maybe moving more cautiously than we should through this process, because it's just not super clear what the parameters are.
So for me, the next three months is saying, "Fence is built. Go buck wild. Let's see what crazy, creative things people can start working on with that newfound freedom." In the next six months, I really want the use of our knowledge to take less intention. I think with how we're building things right now, there's still a great amount of intention involved in saying, "We've built this agent around knowing jurisdictional nuances, but I need to know that it exists, and I need to go to this place."
I want, ultimately, to feel like our knowledge is really deeply embedded in the workflow, and for agents to not just be somewhere that people go visit, but something that's able to surface and embed things in real time. I think that will be the ultimate success. I don't think that's actually realistically gonna happen in 12 months, but at the speed of AI, why not, right?
Chris: Yeah, I like that though. I mean—it's not a zero-sum game. You can move in that direction, towards knowledge being easier to use,
Hillary: Yeah, exactly.
Chris: Which I would argue we've been on that trajectory for the last few years as a community too. I keep calling it the golden era of knowledge management. At KA Connect last year, there was a funny moment, Nicole, where a couple of firms who were brand new to the community—somebody spoke up and was talking about how glad they were to be here, and somebody said, "Man, it's really like you're so lucky—you didn't have to go through the dark ages of knowledge management. You're just coming in after everything's been figured out. Lucky you." And I kind of get it. But I also feel like there's something in what you said, Nicole, and I wanna test this with you, Hillary: the last few years have also been pretty wild.
I think for people who like a lot of change, it's been exciting. But even for me personally, the pace has just gotten pretty insane. And there's something in the way you were saying it, Nicole—it's like we now know enough about the work that needs to be done that we can slow it down and be a little more intentional and not as chaotic, versus whipping around trying to figure out what to do in this environment. I love the way you planned it out—when we go after this process, we know we have to establish this, then we have to do the training, then we have to do the AI piece. Could 2027 be the year we see a little bit of calm restored, and get more intentional about how to build in this new era? I don't know—am I making that up?
Hillary: I've been working with a group of people to put everything into a tracker that lets us measure: is this an idea we should hold onto for when the technology is developed? Is this an idea we could build now? Is this something that's gonna have impact for our client? Is this something that's gonna have impact for us? It allows visibility across multiple teams, for us to see what each other are working on, and there's been iteration that's already occurred at a wild pace from that—that's really cool.
But the other thing is, it's allowed us to hold both things at once. If we only reacted, we'd just make today's processes slightly less annoying. But if we're too focused on the future, we're gonna either build things that people are not ready to adopt, or build things that the technology's just not quite there yet for, and maybe create some distrust that hurts us down the road. So I think you can have the wild and free and also the methodical and organized at the same time. You might think something is the greatest idea one week, and seeing it in the grand scheme of things two weeks later, you're like, eh, it's not really that good of an idea.
Chris: I think what you're also hinting at is—if I understood what you're doing with this tracker—it's also making things visible that we've said this is a good idea, but not today.
Hillary: Yeah, exactly.
Chris: One of my favorite books is a book called The Creative Habit, by Twyla Tharp, who's a choreographer, and the central conceit—the part of her book that stays with me—is that in her mind she's working on choreographing like 25 dances at the same time, but she's actually only working on two. She's getting paid to work on two. So she built a wall in her studio of banker's boxes for all of the future projects she's not actively working on. If she has an idea on one, she pulls it down, puts it in the box, and puts it back up on the wall.
Nicole: Interesting.
Chris: It has a home. She knows we're thinking about the thing, but she doesn't have to work on it today, and she can stay focused on the thing she's working on.
Hillary: And if it's valuable enough, you'll find it and pull it.
Chris: And valuable enough. That led me—I tried doing it with Evernote, and then I ended up using Trello. So I've built her thing in Trello, and I've got projects going out into the future—I've got a home to log all those things. When I decide to work on one, I open it up, and there are all these notes of ideas. Sometimes it's the same note six times over a period of months, that I thought was original each of the six times I wrote it down, right? So I feel like that part of what you're saying, Hillary, is, we can also start to say we don't have to do this all this month. There's time.
Hillary: Yeah. And also, with the speed of AI, you don't have to choose between your reactive and proactive things as much either.
Chris: Yeah.
Hillary: Budget- and time-wise, you used to have to very much choose, because making things was so slow. But now, yes, we can focus, but we can also do a couple of different things at once. We can work on the future and work on the now at the same time, which is cool.
Chris: Yeah.
Nicole: Something that was kind of an aha for us, again—and team structure is a big part of this too—we have a four-person management team: myself, our CEO, our now vice president of practice operations, who we just elevated, and our marketing director. So we've got practice operations, sales and marketing, and CEO and COO. We're all jazzed about AI, all excited about this work, but April, my CEO, and I are the ones who are a bit higher up—everything's exciting, we see the bigger picture down the road about what can be. Whereas my marketing director and our practice lead see that too, but they also have to deal with the day-to-day, and they're managing the teams who are managing the change.
So we had this agreement that, going forward, they can be the gatekeepers. We'll keep on innovating, we'll build the boxes and put them all out there, and they can be the ones to decide how to stage the rollout and change of these.
Chris: So you're the gas and they're the brake, a little.
Nicole: Yeah, exactly. 'Cause they have the best insight into where the team is, and their ability and capacity to absorb the change, facilitate the change—whether they're ready to let us push the gas a little farther, or we need to stop for a little bit. So giving them that permission to say, "I'm gonna stage it out this way," and it's a collaboration. We might push one way, they may push back. But having, feeling like they have ownership and the ability to say, "I wanna stage it out this way, we can't force this too fast," and the insights from the day-to-day work help shape that for us—I don't have that visibility anymore. Even going from 15 to 30 people, I'm losing touch with what's going on day-to-day. So I think setting those roles for ourselves is gonna be really helpful in properly staging out the work going forward.
Chris: I love that. Maybe to start closing on a what's-next theme—you're both aware that we have a Synthesis MCP server now that we just put into private beta. Nicole, I know you've been doing some stuff with it, and I wonder if we can talk about that for a minute, 'cause I think there's something very connected to the conversation we're just having.
Nicole: Yeah, I've been dying for this, so I'm so excited that it's live. We've been talking about how, with AI, we now have access to every knowledge container that we have. With Claude, we're connected to SharePoint, we're connected to Outlook, we're connected to our voice transcripts, we're connected to Slack. So our semi-organized SharePoint folders are accessible, our ethereal conversations are accessible. But until now, all of our standards and all of our most curated, highest-quality shared knowledge amongst the team has been in the Commons, and I've had to copy and paste that in if I wanted to take that as part of the bigger knowledge picture.
This is kind of the missing piece—now I can look at my entire knowledge base all at once. So I think the two highest-value things I'm pulling out of the Commons now, through my central... I don't even know what to call it.
Chris: Yeah, command center.
Nicole: Yes, the command center, exactly. From the command center is our standards. Before, I'd been copying and pasting our standards into things, and that's not good. Now I can just say, follow the standard from the Commons for this. And our project directory—the projects we have, going from Deltek to OpenAsset to the Commons, are our best projects. They all have project summaries, they all have data, and now I can just point to that and grab it all. Whereas prior to that, I was using kind of a poor man's version, where it would search SharePoint, which also has all sorts of data about our projects, but now I have the good stuff.
That's helped me build agents that can traverse multiple of those knowledge bases and the internet to answer questions or get to deliverables, and to have our standards, our projects, and our employee directory really connected to all of that has been the missing link.
Chris: You told me this use case in prepping for KA Connect, and luckily this podcast will come out after KA Connect, so there's no spoiler alert here—that's gonna be in your presentation around updating project descriptions by connecting all those systems. I wonder if you'd be okay to share.
Nicole: Yeah. So our marketing coordinator, as I mentioned earlier, the bane of her existence is trying to get PMs to review project summaries, update them, get the right content in. Even just from a quality perspective, it's fine—we have a boilerplate, a standard for what a good project summary is, how many words it is, the basic content it covers. But it doesn't have the wow—the marketing team can't find those wow things about a project that really sell it. Those are in the PM's brain. And when you ask the PM, "Hey, can you please review this, is this up to date?" they're not thinking about what's really exciting.
So what I'm building with her now is an agent that can pull all of our current project summaries and compare them against project notes in SharePoint and against Slack chatter, and elevate where things might have changed or where some really cool things came up that could take it to the next level. And the final output, for each PM, is just a list of projects that have changed and a redline recommendation. Our marketing coordinator can review that herself and see what she wants to accept, or if she felt like it captured it. And then the last piece is the PMs just do a final review and say, yes, that's right, or not.
So it's capturing things like, for example, an affordable housing site development project that currently just said, "We did this project, we did this site work, the project amount was this much, here's the roles we had." But we discovered through a Slack conversation that we had actually helped save the developer money and found a more sustainable solution to their drainage challenges. That's the wow stuff we want showing up in our project summary—we saved the developer money and found a more sustainable solution—and the current process wasn't even elevating that. We found it through Slack chatter.
Chris: 'Cause they had that conversation in the flow of work.
Nicole: Exactly, exactly. They were just talking about it. So it's streamlining a process, reducing the nagging, making things happen faster. The PMs aren't bothered as much, but it's also surfacing things that the previous way wasn't surfacing, because they happened ethereally.
Chris: So I wanna wrap by asking you both, wearing kind of two hats—one is leaders in your company who have also been building—what advice would you give to other leaders of AEC firms, in terms of why this isn't just something that the innovation team, or the IT team, or some of the emerging professionals who are super good with technology, should be doing?
What have you learned by being in the seats you're in, leading this work, that you think would be useful for other leaders to hear about?
Hillary: I think the first thing is, do not outsource your understanding of this. You do not need to be an AI engineer to build something for yourself, and even if it's something small, you'll learn a great deal in the process. As I said, the more I approach AI problems, the more I learn about the business. And for leaders, how valuable is that? The AI tool itself is very easy to learn—it's really not that complicated, it's essentially the same steps as giving directions to a human being. But what you will learn about your firm firsthand, and where to go, is so valuable.
And I think it allows you to start asking better, harder questions about your firm—about where you're losing energy, and where you can supercharge things. It's a tool to redesign how your work gets done. AI itself is not the strategy.
Chris: Right. I've been saying, I don't think this is AI transformation, this is business transformation powered by AI.
Hillary: 100%, 100%. And I guess the other thing I would say is, as far as the people you should pull in—pull in the people who are already asking, "Why do we do it this way?"
Chris: Yeah.
Hillary: AI doesn't need to be in the conversation. You already know who those people are. And then you have people who've been practicing this craft for 30, 40 years and are super adaptable. You know who those people are too. Pull them into the conversation, because they have the knowledge to transform. And even if AI scares you, the willingness to change is a great indicator.
Chris: I like what you just said—there's something bigger there. Is it possible that we don't need everyone to be as adaptable to change in the organization? Is it possible that if you get a core in the middle that can absorb some of it—be a shock absorber for the organization, take the first hit—and figure out how things settle down, maybe not everyone has to see all of the chaos all of the time. I think that was true for myself.
Nicole: I know who that shock absorber is on my team.
Chris: Yeah. Yeah. Yeah.
Nicole: Yep.
Chris: But even that four-person team you talked about, Nicole, that feels like part of the role of that team, right? You've got people on the team who are like, "No, this is fine, people will be okay with this," and others who are like, "No, this is gonna be really disruptive, we can't pull on this lever right now."
Nicole: Yeah, and it's built on a lot of trust, because we have to trust that they like our ideas, or are excited about us having ideas. They may not like all of them—it's not just like, "Ugh, that's— don't do that, don't do that." They genuinely think it's cool, interesting, see the value, and just not right now. So it is built on trust, to be able to have your big ideas and have people gatekeep them in a respectful, trust-grounded way.
Chris: Hmm. Do you have advice for leaders about business transformation powered by AI, Nicole?
Nicole: Yeah, it's interesting. This conversation has really had me reflect on—at the end of the day, it still comes back to the same business fundamentals we understand if our business is doing well. It's still the same metrics we look to, to tell us: are we meeting our sales goal? Are we hitting our utilization goal? Are we profitable? What are our multipliers? The metrics are the same, and they tell us how we're doing, and the best practices haven't changed. The industry looks to the same groups for best practices in project management, for best practices in proposals.
But for me, I've known those things, but I've been hampered by a couple of things. One is, as an operations person, I'm an inch deep, a mile wide—I know what I'm trying to get to, but I didn't have the skills to develop spreadsheet models, to build reports, or to navigate Deltek in the way I really needed to, to understand how to get the most out of our data, or to do the kind of research that was needed. So it was just too complex for me to feel like I could get from the work we're doing up to driving those metrics, and now I do. These tools are helping me get from where we're trying to go and business fundamentals, and do it faster, better, in the Greenprint way of doing it.
I think that grounding is really helpful for me, as somebody who gets excited by everything, to be like, what's the point? We're trying to do good by our clients, bring in business, keep our employees happy—that's what everything we do should be in service of, and there are decades of work saying what good work looks like, and we should build on that.
Chris: So are you saying—I like this—I think what you're saying is you feel empowered to get answers to the questions you really wanna know, because I don't have to go through someone doing a SQL export, then building a Power BI thing, and then hoping the product is able to actually visualize the things I wanna see. Now you're like, "I've got the data, I've got a question, let me just do my thing."
Hillary: Yeah. The decision has not already passed us by.
Chris: Right. And it isn't three months later, when I finally got the thing built and I missed the window. Yeah. Awesome.
Nicole: We can make decisions faster and get insights faster than ever before, and those are the two things that really hampered us—struggling to get enough information to have insights that can lead us to make decisions. And I finally feel like I have the tools and the insights to be able to do that.
Hillary: Yeah, decision quality, not just speed.
Chris: That's great. Well, Hillary, Nicole, thank you so much for your generosity and time, and for sharing very much in-progress work and thoughts, publicly, on YouTube, Spotify, Apple Podcasts, and wherever you get your podcasts. We appreciate you both.
Hillary: Thanks, Chris.
Chris: You got it.
Nicole: Thanks so much.