Getting AI to Work for You: A Q&A With Document Crunch VP of Product Chris Bruner

Chris Bruner has had an extraordinary career over the last two decades. From his role as an analyst at the widely respected McKinsey & Company to co-founding the educational startup Academic Earth to a host of senior project management roles for companies like Zuora, Vivun and now Document Crunch, where he serves as vice president of product, Bruner has built a reputation for helping companies solve complicated problems through strategic tech application. And recently, we were lucky enough to sit down with him to pick his brain about his career and the impact AI is having on construction, including some best practices for companies looking to adopt the technology as well as what its near-future evolution might look light.

Please tell us a little bit about your background, how you ended up at Document Crunch, and a little bit about your current role there.

My product career has had many chapters, from an early-stage education startup to Zuora’s run up to IPO to working across Corteva’s global suite of digital solutions. The past seven years, I’ve focused on applying AI to business problems; and, of course, what that looks like has changed dramatically during that time. I went from more traditional applications of data science at Corteva, where we helped farmers better predict farming outcomes, to the fast-moving universe of generative AI in an earlier stage startup. I was ultimately drawn to Document Crunch for a number of reasons. There was an immense opportunity to help the people doing some of the most crucial, challenging work to create the built world around us. I also liked the way Document Crunch approached AI products philosophically, keeping the focus on solving business problems front-and-center in order to make humans more effective and successful in their roles.

My role at Document Crunch now includes supporting our product, design, and product marketing teams as we pursue our mission of moving the industry towards zero disputes and stronger relationships.

AI is proliferating throughout the AEC industry, making its way into workflows and changing the way people build and do business. In your experience, what are the biggest hurdles companies face in implementing those technologies such as not to disrupt the very workflows they aim to improve?

Two things come to mind for me.

First, it’s so valuable to build your mental model of where these AI technologies can be effective. They can feel magical at times and at others still so limited. So, there can be a temptation both to over rely on it and to reject it. But, when you understand the space between those extremes, there’s an immense opportunity to benefit. This starts with leaders at firms framing the opportunity (e.g., “This tool can flag risky clauses in minutes instead of hours, but it still requires your expertise to interpret and act”). Then, as team members experience useful applications for themselves, they can build their own mental models of how much autonomy they can give these technologies and where they can benefit from them. This is also where having domain-specific AI solutions can be so valuable because these move users much faster to a moment where things “just work.”

AI technology is advancing so swiftly—faster than any foundational technology I’ve seen in my career. Building an AI solution may work, but whatever approach you took may be obsolete within months.

Second, change management is non-trivial. We all now live in a world where, however much we have honed our craft over the past many years, we have to rethink how we do things. Many of the ways that we work can now be transformed for the better. However, to experience the benefit, we first have to give up ways of working that have long been successful and adopt new, better ways of working. Initial training helps, but we see the most success from companies that build in mechanisms for ongoing coaching, workflow refinement, and reinforcement. Using networks of peers can also be very powerful here. Pilot groups of energetic early adopters light the way for the middle majority of users to follow. This also creates opportunities for innovators to have a positive impact on their peers. Recognizing these folks and their efforts, positioning them to share their wins, and elevating them as champions is both great for their growth and great for the organization.

Lastly, it’s valuable to simply dive in. Sometimes there is a temptation to wait for the next project to start, but that introduces dependencies. Teams are able to find ways to benefit from work already in flight, and that starts the learning flywheel more quickly.

What are some best practices or good questions for companies to implement/ask themselves when adopting AI for the first time?

It’s important to have executive sponsorship and invest strategically, but it’s not necessary to revolutionize everything at once. The teams that succeed often start with select, high-pain workflows tied to company initiatives and goals (ie, reducing risk, reducing rework, etc). They prove value quickly and then expand. Momentum beats magnitude.

I also encourage companies to think about how they start but also about how they sustain. AI technology is advancing so swiftly—faster than any foundational technology I’ve seen in my career. Building an AI solution may work, but whatever approach you took may be obsolete within months. Have you also built the capability to keep evolving that solution to keep pace or partnered with a vendor who is going to keep you at the frontier? As powerful as tools like Claude, Gemini, and ChatGPT are, once you get into the details, you see how much you need to do to create successful business applications of those solutions. How is project knowledge pre-processed and modeled in a way that foundation models can readily make sense of it? How do you encapsulate unique industry and company procedural knowledge in a way that helps foundation models graduate from generalists to useful professionals? In an industry with such enormous amounts of content to process like construction, how do you engineer context to overcome context rot limitations? How do you design the human workflow on top of that so humans can supervise and direct AI agent efforts efficiently?

Finally, I know that industry peer groups have been helpful for many when it comes to AI. Customers often share with each other what solutions they have implemented and come to trust, and I think that is a great way to de-risk getting started.

What is the next phase of AI in AEC, so far as you see it?

One of the things that I am keen to see is how AI solutions become repositories of expertise. Today, it’s not easy to ensure that hard-won lessons learned from one project roll over to the next. Even if you are disciplined about documenting each of those, how easy is it to ensure each team member calls the right one to mind at the moment of need? That’s where AI provides such a useful vehicle. You are able to capture those in the form of standards and then AI is able to map the right past learning onto the present moment. Many companies are setting themselves up for this today, and I expect that this will have a compounding effect over time.

there’s a temptation for big companies to hang onto their data more tightly amid this AI transition. The belief is that this will protect their market position in a world of small, fast-moving new entrants. But I don’t think there’s a lot of evidence that data from traditional software solutions is the long-term competitive moat.

When I think further out, it’s fascinating to see where some of the leading AI thinkers have now turned their attention. Fei-Fei Li, a Stanford professor and one of the key AI pioneers, argues that one of the next major horizons in AI development is the evolution of world models. The idea is that this helps move us from applications where we’re predicting text to applications that have spatial intelligence and are able to perceive and reason about three-dimensional spaces.

One of the things in the works at Document Crunch is better surfacing risk in construction drawings. There are a number of ways we can be helpful immediately there, but there are other places where I think it’ll be important to evolve the technology further. Designs may be vetted upfront for clash detection, but then there’s a huge amount of detail that’s added once that’s translated into what can be built. How do you trace the implications of a shop drawing for the clearance requirements in a space? How do you understand how one change in a drawing set cascades through the work of all the trades involved in an effort? These kinds of things require a great deal of painstaking effort today, and they’re often addressed amid short lead times. But, if you could extract the geometry of a building and the relationships between its components, could you start to understand how things need to relate to each other in a physical space and identify potential conflicts and inconsistencies much faster and more efficiently?

Beyond AI, what is the tech or innovations in general that have you most excited to see used in AEC in the near- to mid-future?

It’s a simple thing, but amid this change, I especially appreciate the technology players who continue to invest in open ecosystems. I saw this in my previous role in the world of GTM software, but there’s a temptation for big companies to hang onto their data more tightly amid this AI transition. The belief is that this will protect their market position in a world of small, fast-moving new entrants. But I don’t think there’s a lot of evidence that data from traditional software solutions is the long-term competitive moat. And customers benefit much more from connected and interoperable ecosystems. I feel fortunate that, at Document Crunch, we have benefited from working with a strong group of larger partners who are moving in that direction. That feels like a great thing for the industry, at a moment when there’s a lot of value to be gained.