Rebuilding Synthesis: Christopher Parsons on AI, Community & the Future of AEC Knowledge

Hi everyone. I'm super excited to host Christopher Parsons, founder and CEO of Knowledge Architecture. And I can't wait to see all the AI magic feature you just released and the product you have built. So thanks a lot for being on the show.

It's my pleasure. I've been appreciating what you've been doing on LinkedIn, so keep it up.

Yeah. So Christopher, right after graduation to this date, if I were to ask three highlights of your career, what would those be?

I think getting into architecture at all. So I was a history major and graduated in the middle of the dot com boom. So I don't know that all your listeners know what that was, but it was a crazy explosion of technology around kind of the late 90s and early 2000s. So I was going to go to law school, but because I graduated during that timeframe, I just got swept into technology and worked for a technology consulting firm during that boom. But then it eventually crashed and like all the dot com stuff fell apart. And then 911 happened and so the job market in San Francisco was bleak and so I found a job in an architecture firm as kind of like an IT manager to kind of ride out that recession. My plan was to go back tech, but I fell in love with architecture, ended up staying at that firm for five years, became their IT director, went to a larger firm, became their Chief Information officer, did that for three years, and then started Knowledge Architecture, which would definitely be my second highlight was just building a company and it was kind of a fork in the road for my career. I loved doing knowledge management in aec. So I could either go to a much larger firm and be their like chief Knowledge Officer or something, or start a company and work with kind of more small and mid sized firms, which is kind of the firm size I grew up in. And so that's what I decided to do. So getting started was a huge highlight. There are so many, it's like hard to edit. Now that I've used two, I'm down to one. I'm going to tie like one is just building this amazing team that I've had the pleasure to work with at our company and then the second and the community that we work with. But we just did this massive replatforming of the product into what we call synthesis. Six, moving it to the cloud, dramatically simplifying it. Now we're adding in AI functionality, we're adding in learning management functionality. So we've been in business for 16 years, but in some ways it feels like we're a startup all over again. There's all this new opportunity with AI and beyond, and we've just got that kind of experience from doing this for this long. But also we're kind of looking at things with fresh eyes and kind of a beginner's mind. And it's just. This is kind of maybe my favorite year of Knowledge Architecture, like where we are right now.

It's just super fun and like, it's, it's incredible. You, you have the same like enthusiasm of running the company so long and with handling new products with, with such great excitement. So, Chris, from POC of Knowledge Architecture to production, do you have any pivot story or anything you'd like to share how you grew and built a company?

Yeah, I'm not a business major and I did not come from a family of entrepreneurs, so I was figuring it out as I went along. And when I decided to start Knowledge Architecture, it was during the housing crisis in 2009. And so what I did is I went part time and I went down to three days a week. And in my extra days per week and nights and weekends, I was experimenting, trying to figure out what Knowledge Architecture could be. I didn't know if it was going to be a consulting firm. Like, I kind of thought there was one version of this. Like I'd have four clients and I would do one day a week at each one and help them with knowledge management. There was another version we were building like just integration software or reporting software. Intranet was one of the things we were thinking about building. And so for about 10 months I was working two jobs. I was working my old job at part time and then I was building ka. And through that period we were just iterating and experimenting. We're taking on projects, we were trying different things. And after that amount of time, it was clear the Internet was the thing we wanted to build from Synthesis, which is our intranet platform for AEC firms. So just that kind of experimentation in the beginning really let us understand not only what was going to work from a business perspective and provide kind of recurring revenue and just a stable foundation, but also what we liked doing the most. So I would, I would definitely encourage folks to experiment like it's, it's a great way, especially if you can do it while you're still earning an income. That's the way to do it, in my view.

Amazing. And you know, people talk about, for AEC founders, the distribution playbook, how to get customers and how to build a business in a sustainable way. Any lessons learned you would like to Share.

Wow. I mean part of it is I did come up as an AECIT leader and so I did have a bit of a network already and they had kind of over the years I was part of some different roundtables and involved in community and they'd kind of seen what I'd built and we're like, dude, if you ever want to start something, let me know. Like, I like what you're doing. I had kind of built intranets at both of those firms. It was a lot harder than it should have been. And so like people kind of saw that as a potential thing for us to do. So part of it was that initial network, but I had to go much bigger than that to build a company. And a lot of speaking, so did a lot of thought leadership, a lot of talking about what does even knowledge management mean? What does it mean in aec? And that was speaking at other events. But then we also launched a conference in our first year of business called K Connect to kind of talk about knowledge management AEC and we expected like 15 or 20 people to come and 80 came to the first one and it just became this kind of thing. We recorded all the talks, we put them online for free. And so my kind of advice there is to kind of create more value than you capture and lead and share. If you've got kind of research or opinions or good examples from clients, like tell a story. It's not just about the technology. People need to see how that technology fits into their world, understands how it solves the real problems. And I think that that was really useful for Knowledge Architect kind of take that approach and community too. Just like learning. I'm not like I'm not the guru of knowledge management in ac. I don't know everything. I've learned more from our community than I know myself. And so like empowering others to share the pieces and parts that they know and like building community, I think that's always a great way to get to get a company going.

Amazing. Yeah. Like I often hear in aec, to run like a startup, relationships are quite important. And through community and constant value offering, I'm sure there are lots of ups and downs in the startup. Any particular incident where, where you had to make a tough decisions and the startup was Louis and what, what we can learn from that wisdom.

Yeah, I mean I think the biggest kind of defining moment for us was around kind of 2016, 2017 when we initially. This is actually kind of answering two of your questions at once. One of the we didn't raise money so we're a bootstrap company. And one of the ways we made that work is we built our solution on top of Microsoft SharePoint. So that's what I had done before. So I had experience with it. SharePoint has improved quite a bit since 2009, but back then, it was very hard to build anything on it. It was not really a product. It was kind of a series of capabilities and features, and you kind of had to stitch together a solution out of it. And so that's kind of how we got the company built, is that we. There was kind of generic SharePoint which most people were trying to run, and then we added an AEC layer over the top of it, right? So it integrated with systems like Deltec and Uninet and Open Asset, new formula, like, systems. People cared about it, kind of talked about employees and projects because we understand the business, we cared about the. We made the search really work. We added analytics, we made it mobile. So it was this kind of like, increasingly thick layer around SharePoint. But that was SharePoint on prem. And then Microsoft started very clearly signaling the future of SharePoint is in the cloud. It's SharePoint Online. We would have had to rewrite the product to kind of go on to SharePoint Online. And so the fork in the road for us was, do we rewrite synthesis and take it to SharePoint Online, or do we just start over and build our product from the ground up, but control all the code? And that was a really. We didn't know what our clients would think. The degree of difficulty technically seemed really hard. And so we spent some time thinking about it. And eventually I was talking to my wife, who's actually my design partner on the software, and just kind of talking through the decision. I'm like, why don't we just go to clients and ask, like, just talk it through with them, like, get their feedback. And she's like, are you sure? Like, like. So we kept talking and eventually, like, she pulled out of me the statement that, well, I don't want to work in the company that's built on top of SharePoint anymore. Like, it's very frustrating to be built on top of another platform. It's constantly moving underneath you. You want to fix things for your clients, but you can't because it's not your code. And I'm like, I don't want to do that anymore. She goes, great, then what's the decision? You know what you need to do? You need to start over and replatform the product and build it from scratch. And so that's what we did. And it took about five years. It took the first half of my 40s to build this product. But it was the best decision we ever made. Like, clients are super excited about it. It's easier to support, it's a simpler product, but it's more powerful. We can truly control the code. And so it was a hard fork in the road, but like, I really dug deep. And it wasn't so much about the fear of what clients are going to say or do. It was really about what kind of world do you want to live in, what kind of life do you want to have? How do you want to imagine the next 5, 10, 20 years of your career? And when I looked at it that way, it was clear I knew which path to go. It was, you know, we still. So what I ended up doing was a listening tour with all of our clients. So I still talked to all of over a period of about six months saying like, look, here's where we're going. What are you excited about? What are your concerns? Like, what are the opportunities in this new platform? What are you going to miss from SharePoint? Like, we still did a big listening tour with all of our clients, but it wasn't like, which way should we go? It's like, we're going this way. How can we make this great for you? Switch sides. You're on the left now.

Yeah, I'll just merge both clips. No worries. So I think it started the recording already. Chris, can you walk us through knowledge Architecture, what the product does, how it benefit firms?

Yeah. Do you want me to show it or do you want me to talk about it?

Whatever is more convenient.

Maybe I'll start by talking about it and then maybe I can show a few things at a high level. We think about kind of three main things we're trying to help firms with. There's kind of building a knowledge base for the company, right? This is anything from standards and policies and best practices and you know, who are all the employees in the company and what do they know and their skills and expertise. What are the projects we've built, who is the project manager? How much did it cost? Like just like really all that stuff, that information that people need to do their jobs. So we kind of create that searchable knowledge base of firm wide knowledge for the company. So that's kind of part one. Part two is in internal communications. So, you know, allowing people to post and share, you know, new projects, new strategy work, new tools, you know, innovations that have happened. Recent wins of New projects, like all the kind of stuff that's going on in a dynamic AEC firm. There's an internal communications layer and then third is about learning and development. So we've got tools to make it easy to ask questions and get answers, share lessons learned. We have video libraries in the product and we're now introducing a learning management component into the product as well to allow you to create courses and assign courses and track progress and all those kind of things. So when you kind of look at it kind of like the knowledge base, internal communications and learning and development, it's really a knowledge management platform for AEC firms and it's, it's really fun to build.

It's amazing. I would like to see how it works.

Sure, yeah, let's do that. I'll share. Okay, that looks right. How are we doing? Are we looking at synthesis?

Perfect. Cool.

So this is the homepage. This is just a demo environment. So it's customizable in terms of firms, brands and all that kind of things. But I think this is pretty representative of what you would typically see. So, you know, I mentioned the internal communications layer. So I'm kind of seeing here on the homepage, you know, different posts, you know, that people have written. I'll switch it into this view and then you can see that they've got comments and they've got likes. This is a video, so you've got native video directly in the platform, which, you know, is. We automatically transcribe all the video. We've got an AEC specific video transcription model. So it understands all the terminology, you know, of AEC firms like Revit and Deltec and all those, you know, ashrae, all those kind of technical terms. This helps us get more accurate transcriptions on the knowledge. You know, I'm kind of going across the top navigation. These are what we call communities in synthesis. And so this can be anything from, you know, a standards library to we've got a feature called Guides which lets you kind of set up, you know, kind of things that used to be a hundred page PDFs about how the firm works. You can build guides, you know, in synthesis. So from a design technology perspective, you know, I'm going to like a Revit page. I can see all of our Revit resources. I can see who the key points of contact for Revit are, recent posts about Revit, some training videos on Revit. I can go into a standard guide and understand, you know, how we approach container files, you know, that kind of thing. And then from a learning, I'm Just going at a very high level from like a learning perspective. Like this is like a software training library. And so it's easy to kind of add videos and make all that searchable. So that's, that's kind of like a high level of synthesis. I mentioned employees and projects before. So we do have an employee directory. This is pulling data in from systems folks already have like Azure Active Directory or Deltec Uninet Open Asset, those kind of tools. You can kind of filter and search. And then if I go to, you know, an employee, so I'll go to our CEO here in our fictitious company, Davison. So it's kind of like an internal LinkedIn profile, right. I can understand what she's worked on in terms of projects. I can see her bio, her skills, licenses. If I go back, I'll go to the project directory too because it works similarly. Same kind of idea, right? So I've got this directory of all these projects. I can filter and sort by different, by different items. I can see all of our projects, you know, kind of on a map. And then if I search on one of, if I, I'll just go to one of those projects, I'll go to Canal Street. I've got images for that project. I can see who the project team was, data points like construction costs, square footage, where it was located, awards and so on and so forth project descriptions. So as I mentioned, like this is really the kind of firm wide knowledge base to understand how we work, who works here, what we've worked on. It's got the internal communications piece and then it's got learning. So that's kind of a high level. And then the search piece is really what we've been pushing on really hard over the last several months since we went into public beta with AI search. So I can ask a question now that you know all the content that's in synthesis, I can ask a question like, why is my revit model slow? And our AI search is going to go across all of the documents, videos, pages, everything in synthesis, send it to AI to then generate an AI search summary. Maybe my question should be why is this search slow? This is not a usual performance for us. There we go. So what I get back is the AI generated search summary. It is giving citations and references to where it drew that knowledge. So if I click on model upkeep, I can go understand like, okay, it pulled this information from this model upkeep page. I've got filters across the top so I can kind of drill down on a certain piece of Content I can ask follow up questions. So for example, like in part two it says manage linked files. So what if, I don't know, I'm new to the firm, like what are, you know, linked? And then it's going to generate just an answer, you know, for you there. So that's kind of the basics of AI search. And I mentioned video before. So if I can say like what's our approach to math? Timbra. And then I get a search summary, you can see that it's searching across videos and documents. We actually have a connection with Zendesk, so it pulls in some Zendesk resources. But let's say that I'm just trying to write something from like a marketing perspective. I can filter that and just pull things out of, you know, one of our marketing document libraries. So I'm just pulling firm boilerplate. It's going to regenerate the summary and then boom, you've got that. So this is being used, you know, by design technology and IT teams to kind of provide better end user support so that people can just ask questions here and get answers based on documentation and videos and all that kind of stuff. It's being used by marketing teams to help them speed up the process of writing proposals or kind of pursuing interviews. Every employee can ask questions about their vacation policy or the compensation policy or just anything you can think of. You've kind of put all of your firm wide knowledge here and now people can just search and talk to it like, you know, like they're using ChatGPT or like they're kind of just searching naturally. So that's kind of synthesis at a high level.

This is so incredible.

Like thank you.

And I can totally see like to make AI useful in businesses and it has to work with the data and someone has to do like the hard work of aggregating all data in one source and synthesis. Already did that. And now AI magic with searching and like human search through multiple folders might take like minutes and you might not even find the right information. Here you're getting aggregated with source table. It's amazing.

Yeah, it's really, it's, it's interesting what you said because like, you know, whenever you build a new product there's like things you think it's going to be good at and then these things emerge and one of the things that's kind of shown up is this idea that like traditional keyword search will give you links to things, right. That you then have to go read. And it could be that to answer your question, you actually need pieces and parts from like five or six different things. And so we've kind of synthesized that, no pun intended, into kind of like one answer. So it's kind of this synthetic aggregated answer. And like that was an emergent property of the system that we weren't, we weren't exactly building for. But it's been pretty powerful because one of our clients I was talking to earlier this week said, you know, that is saving us from have. It's basically dynamically generating pages we would have had to create on our own. But it's kind of infinitely scalable because you can't anticipate all the questions that people are ever going to have. So what you want to do is put your best knowledge forward and let people just ask questions of it. Like you just don't know where it's coming from. So it's, it's been fun to work on.

Nice. And just gave me an, like an idea if I'm, let's say aggregating multiple proposals or like what are the strategies one can build on top of this? Like automatic proposal creation tool for new projects. Or like, because I've learned the best practices in, in the documents I've shared. And AI can be autocomplete tool for.

Yeah, I think so. I think we're not all the way at automating proposals yet. I think that's a, that's a, that's a little bit further out in the future. I don't know if that's what we'll, I mean, we'll see. But I think what we have done, working with our kind of, you know, the marketing directors and CMOs and our clients is kind of look at their whole process and kind of target the parts of them that are super labor intensive and hard and just find. So if you have a question like, you know, what are all of our projects that have floor to ceiling glass windows? Something like that. Synthesis can bring back the list of all those projects and it can have like descriptions of each of those projects. That previous task would have taken somebody hours to like go through every old project one by one and figure out, read all the text, find out if it has floors. Like we're doing that in seconds. So the head start you get in terms of gathering that information and then you can give it a prompt to say like, please write this. You know, when I did, I did the question before, what's our approach to match Timber? But then we can do things like please write this in a narrative, please write this as an faq. Please Write this, you know, like you can kind of give yourself a good jumping off point to then kind of take it into your proposal. So they're still going to do the kind of the formatting and get the template right. Like we haven't tried to build any of that stuff but that first step of like gathering the information and having like kind of a draft response, like we have dramatically simplified that. So that's pretty exciting. But who knows what comes next? You know, like we're working directly in partnership with clients and they're going to tell us what the next most important thing to build is.

And do you envision like plugin system where other developers can build on top of synthesis?

We're open to it. We, you know, the way that we kind of think about roadmap is it's very community driven and we stay very close to our clients and they tell, they make it very clear what the next most important thing for us to build is. So if that's what they make clear to us, we will do it. Right now they've made some other things clear to us around AI and so that's kind of what we're focused on, on building those things.

Any advice for people who want to build AI products in AEC from your knowledge, like what are some opportunities to build challenges to overcome?

Wow, that's a good question. I think I would start from, I would try and take a pretty long view and try and to the extent possible, try and be three years or four years in the future and look back on today and try and understand is this opportunity that you're perceiving or the problem you're trying to solve still going to be is it a durable enough problem? Because the AI technology, you know, the underlying models are getting better. Right. And so the question is all you're doing like, so I just, let me just use our product as an example. The secret sauce in building AI search is not the LLM. It is doing all the retrieval of the information, understanding what the most port, building a UI that people want to put information to in the first place, then being able to understand what the query is, you know, based on domain knowledge, based on user intent, then understanding what information to pull back, ranking that information, sending it to the. So it's like the AI piece is like I don't know at the end, 5, 10% of the solution. It's all the other stuff that you have to do, the deep understanding of workflow and data in the domain for your, for your client, like that's what's going to make this stuff durable. So I think it's like find a problem that you're passionate that you will work for me, 16 years and more. But for, you know, somebody, 5 years, 10 years, 15 years, like find a deep enough problem that's exciting, not something that seems like a little quick opportunity in this moment. And that's because, I mean, you know, work is hard and if you find a problem that you're passionate about that's valuable enough that can power you through the ups and downs. And so I don't know, that's. That's kind of my take. Like the secret of AI is it's just kind of a technology. I think it's kind of like when databases became a big thing, or the cloud became a big thing, or mobile became a big thing, it was hot on its own sake. But eventually you had to build applications that delivered value for users. And there's a lot more to building applications than AI.

When you were talking word, the video transcription service you made where there are like AC specific terms so there are always like opportunities where even for construction drawing like out of the box retrieval doesn't work because there are multiple details. So there's opportunity of training AC specific detailed cropping retrieval to some AI can get those image. So identifying opportunities of innovation.

Yeah, Chris, did you want me to talk about the transcription thing or you were just saying there's more opportunities?

Yeah, I was just highlighting like the kind of innovation you did. Anything which we didn't discuss in the interview and would like to share that on an ending note.

Maybe two things if I could. One would be actually I wanted to pick up on that video transcription model because we're building another AEC specific model for the nerds in the room. We're building an embeddings model which enables kind of retrieval log metageneration for the non nerds in the room. We're building a tool that will understand people's search queries and your content, but understand it from a more deeply AEC specific context and be able to find the right results. So it'll understand things like standards are similar to guidelines. It will understand that like, you know, frit is kind of related to the concept of like bird safe windows, like just all that kind of nuance to make search possible because people are going to come with different words for the same thing. And so we're building that the way we're building that is we built a. We have a community AI program within our client base so people can opt in to help us build these AEC specific models. So we train on their content, but it's only used for retrieval, like I mentioned before, and video transcription, you can't search each other's content or anything like that. So that's been great. We've got something like 75% of our clients have opted into that to allow us to kind of build with their data, which is amazing. Like, that's partly the trust we've built, you know, through the years. And so that's, that's, that, that, that kind of like AEC specific AI thing I think is going to turn out to be really cool. And the last thing I wanted to mention is we have an annual conference for knowledge management called K Connect. It's held in Sundance, Utah. It's coming up in August. And this is a community. If you're kind of like, if you're into this, listening to this and it's like knowledge management sounds interesting. Working with data sounds interesting. Like these are your people. Like they are knowledge managers, nerds, like peep data. People like that want to get into that kind of thing. And so that's a, that's a fun thing to check out.

Amazing. I'll add that in the description. Description notes. Thanks a lot, Christopher.

My pleasure.

Amazing product and like heads off. It's inspiring for your perseverance you had and like the product you built.

Thank you. And thank you for the work you're doing to connect a bunch of people. I've seen how many people are following you and into your stuff and that's kudos. Like, it's really good the work that you're doing.

Thank you.

All right, take care of.