How AEC Firms Are Managing Outdated Knowledge in the Age of AI Search

When so AI search has been really interesting because it is provided so much more visibility into content that just kind of hangs out. And in some cases that content hanging out is great because it is the only information about that particular subject that exists on our Internet. But in some cases it's just hanging out because it is no longer accurate and hasn't been for two years. Our intranet is 11 years old. Some of the things that we put into it 11 years ago were accurate and correct at the time, but are no longer the case. Right. So I would say at the moment, our strategy is we have articulated to our folks, so if you see something weird or old, trust but verify, which is really good advice in the world of AI, and bring it up if you see something that shouldn't be there. And we will address it at that time. I am hopeful that at some point we can develop more sophisticated tools to help do that proactively because it would be better if no one was ever presented with information that was out of date or inaccurate from our Internet. But as much as we use it as a tool for communications, it just doesn't seem like that's doable right now. So we know that that's a problem and we struggle with that as much as anyone else does. And at some point we're going to figure out a way to solve that problem, and we do. We're going to let everyone else know in case it's helpful for them too.

Some of the things in our community, people do, they do a little. And correct me if I'm wrong if you guys are doing this, Jim, but like, they'll just set up a rotation schedule to go through the different sections of their intranet and make sure they hit everything twice a year. You know, kind of like a. That's. That's a way to do it. There are some technologies that we could put in play. I think that one of the things that's so hard about it is it's not always clear when something's out of date. It's not how old it is because there's things that were written 11 years ago that are still as true as they were. And something six months old or three months old could also be out of date. So it's an interesting open awareness. So, like as many tools and processes you put in place, some of it's just like culture and awareness and habit of, like, knowing that, like, if I'm making, if I'm making a change, we need to make sure that it's updated in the knowledge base and that gets tricky to roll through an entire culture or the person that added the thing is no longer at the firm. And so it's a. It's. I think that Jim's point around AI search has just helped because you can start seeing the people that some things are inaccurate and don't really matter because no one's going to see it anyway, but then it will. AI searches have this way of kind of reactively. I think the kind of more review process would be proactive. The reactive way of being like, no, this is stuff people are seeing and they're coming across this on a regular basis and it's out of dip. It helps you prioritize the kind of more reactive knowledge management side in addition to the proactive strategy, if that makes sense.

Yeah, but we have videos of town halls that we've run or quarterly meetings that we ran in 2018, 2017. And some of the stuff that's in there is great. But a couple of things we said in there is not our current practice. Yeah, but going into an hour and a half long video from eight years ago and chopping out the piece that is no longer relevant is a level of effort that we have not managed to build up the interest in solving.

Yeah, and on top of that, Jim, like, and this is something that's come up at Shepley, but other places too, is even if it's inaccurate, sometimes people want to preserve it as an artifact of culture. Right. So it's like sometimes inaccurate's fine because this is who we were back in 2014 when we were doing this thing. And we don't want that to change, but we also don't want people to get the wrong information. So there's all this like, stuff in AI around waiting and kind of of biasing towards more recent concept. You know, there's like a bunch of different strategies to take, but some of that is like, what was the thing you said in your presentation, Jess? It's not hoarding if it's organized. And so some of it is like, if we're going to keep more stuff rather than being ruthless and just deleting everything that's out of date, like, how do we trust but verify, I guess is part of that, Jim. It's like, just take everything with it. You have to. People know, need to know to like, question what they're seeing.

The waiting is an interesting thing because that's a strategy. We move towards more than removing stuff. If something's really bad, we're going to remove it. But it's also possible to just bury it down. So, for example, we had problems with file path links. This is going to get really crunchy and technical, but all of our documentation. I know, right? All our documentation about file path links refers to new forma. So if you ask a question about it, then, like, you're like, oh, we're not using new forma, therefore this is not a problem anymore. And that's not true. So what? We built an explicit page with explicit language that we knew the AI would like to pick up, and so now it doesn't refer to new forma because that newer page doesn't refer to it at all and has more context outside of it.

So we're finding ways to mitigate the problem. But I think for me, the most important strategy we are running is trying to build discerning individuals in the firm, because discernment, I think, is going to be a really crucial skill moving forward in the world of AI. And if we can help people understand how to deal with the cognitive dissonance of this is a true fact, but check to make sure that that true fact is still true.

Yeah.

Feels like that's the sort of thing we're going to have to get better at as a species in order to deal with the onslaught of information that's coming at us beyond the Internet.