In this webinar, Christopher Parsons shares practical insights gathered from a private beta of Synthesis Knowledge Agents aimed at architecture, engineering, and construction (AEC) firms. He outlines a decision framework that asks whether to spend an hour performing a task manually or building an agent that can automate the task repeatedly. He stresses starting with business problems—identifying knowledge gaps—rather than jumping to technology, and recommends embedding oneself in workflows to surface high‑value use cases. A repeatable pattern emerged: pair technical agent builders with subject‑matter experts to interview, extract critical knowledge, and co‑create agents. Codifying firm knowledge into agents forces firms to define “good” outcomes, speeds work, reduces repetitive questions, and empowers junior staff. Parsons discusses when to distill knowledge into concise best‑practice documents (e.g., NDA review) versus when to give agents access to large data sets for long‑tail search (e.g., precedent discovery). He highlights that giving agents a clear name, icon, and use‑case description dramatically improves adoption and AI literacy across the organization. Finally, he argues that having senior leaders personally build agents creates a conversion experience that reshapes their perception of AI, builds credibility, unlocks resources, and accelerates firm‑wide transformation.
Chapters
So one of the things I talked about in this webinar description is I wanted to talk a little bit about what we're learning from the beta. So a few kind of like quick hitters. So identifying agent opportunities in this category, we've got do the task versus build the agent. These are all direct quotes from clients. So now, like kind of the change after working with agents is when I'm starting to do a thing, I ask, should I spend an hour doing this task or should I spend that same hour building the agent that can do it? And then I've got an agent that can do it over and over again. Or starting with knowledge gaps, not technology. And so, and this is really kind of getting at business problems or challenges first. And so spending time with people, understanding where they feel blocked, they don't have access to the knowledge they need, and that helps you identify high-quality, uh, use cases for pursuing agents. This is a classic knowledge management trick, is to insert yourself in workflows, whether you're asked sometimes or not, and just understand like how people are working, watch their flow of work and realize that, oh my gosh, if we could capture this knowledge and put it in the flow of work with an agent for somebody, we could really streamline their work and make it a lot more joyful. From a building effective agents perspective, one thing, a pattern we've seen over and over again is that we're pairing people that are agent builders with subject matter experts to work side by side, help do the interview, right? To maybe pull out some of that critical knowledge, give them a draft of what they've learned. Help translate their vision and their problem into actions and all the technical parts that make an agent, and then pair program and build these knowledge agents together. As a result, we've seen agents forcing firms to clearly define what good looks like. They see now the value. This is the return. If we document our knowledge and if we build this into an agent, we can move faster, we can reduce the amount of questions that people have to offer, we can empower emerging professionals to take work on earlier in their career. There's all these benefits, they all depend on codifying the firm's knowledge and processes, and so codifying the firm way. And so agents are acting as a catalyst, and we're seeing more and more knowledge management activity as a result. Um, one of the things that comes up, and, and, uh, I will touch on this briefly, we can talk about it a little more in the discussion afterwards, is as people have worked through agents, they're trying to figure out, so do we distill something into a best practice document, Or do we give it a bunch of PDFs and let people look through them? And I showed examples of both, right? In the highly repeatable process agents, like the NDA example, we wanted our best practices written down. We didn't wanna point the agent at all the NDAs we reviewed after the last 2 years and, you know, sort it out. Good luck. Right? So that, that is worth distilling and making a best practice document. On the marketing example I showed though, like for precedent discovery, for example, We wanted it to do long-tail search and then giving it lots of projects, like whatever the use case is, to search through. And so there's this thing that you'll learn and, and kind of adapt to figure out, like, when do we take the time to distill the best practice and when do we just give it access to a lot of data? Um, 2 more things on scaling organizational impact. One thing that we have heard over and over and over again in the beta is that the agent format itself, meaning it's got a name, it's got a clearly defined use case, it's got a cute icon and a description. Helps people understand how AI can help them. So instead of saying to somebody like, hey, you have a license to X AI software, go do useful things, we can be like, look, we built an NDA review assistant. When you're reviewing NDAs, use this thing, or use this thing for finding design precedents, or use this thing for staffing your projects, or use this thing for getting sustainability advice. So this kind of way of combining that grounding knowledge plus a use case plus instructions helps make this much more democratized for other people and raises people's literacy. And I've kind of said the same thing here. It's slightly different, but this is really bringing more people into the fold of using AI who maybe never would have built on their own. But once you show them something valuable, they'll start wanting to use it. So, at a high level, you know, that's what we've learned in the private beta. But there's one more thing I wrote about this earlier this month. This is in the June issue of the Smarter by Design newsletter, that leaders should be building knowledge agents. Not every firm leader, but you need some leaders in there. They have the highest visibility into the firm's high-value opportunities, right? Building an agent changes how leaders think, not if they watch a demo, but if they actually personally build one and make it useful, they then start seeing agent opportunities everywhere in recurring workflows, reviews, if they're in training, they're just now able to see the world through a different lens. So having them have that kind of conversion epiphany experience is really, really useful. Um, and then they bring credibility to AI transformation. They can inspire other leaders. They can unlock resources to do that knowledge capture that's necessary to power agents. And it will just accelerate your firm's, you know, evolution of business transformation if you get leaders involved and not just watching, not just seeing demos, but personally building agents. I'm telling you, that will make a huge difference in the way your firm's adopting AI.
