Why do most AI pilots stall before they reach enterprise scale?

[00:00] Why AI pilots stall: the hidden knowledge problem

JP: Most AI pilots work, and then they stall before they even reach enterprise scale. One big reason is hidden knowledge. The expertise scattered across your company that nobody wrote down. Everyone holds a piece of it, and nobody sees the whole picture. If your people can't see it, neither can AI. Hola, I'm Joe Pardavila from Forbes Books, and this is geniant, The Road to Project Zero. My guest is Chris Byler, Principal of Applied AI at geniant, an AI consulting, design, and engineering partner for the enterprise. Chris is helping build Project Zero, geniant's platform for taking AI from successful pilot to enterprise scale. Today we're covering where hidden knowledge come from, how knowledge graphs keep AI aligned with your business, and how to keep people in charge of the work. Chris Byler, welcome. How are you, my friend? Doing good. Excellent, good to hear that. Chris, I want to start with the name Project Zero because outside of it being a really cool name, like kudos to whoever came up with the Project Zero name,

[00:56] What makes Project Zero different (and where the name came from)

JP: But how does Project Zero differ to how other companies are using AI right now? Because as you know, Chris, everyone's using AI in some shape, form or other. Tell me the differentiator with Project Zero. Yeah, sure.

CB: The Project Zero is a fascinating concept. It's kind of a little bit of thought experiment, a little bit of this and that, but what we see as Project Zero is the culmination and direction of where AI is headed in the industry for the sake of currently we're looking at it toward enterprises and large organizations, of course. But Project Zero originally started with the name of the last project will ever deliver for you is kind of where we originally originated with it. And so obviously we needed to rebrand that a little bit better. So our guys over in London, they came in and used their amazing design skills to come up with the name. But so Project Zero is essentially a platform that brings together experts to be able to work alongside AI and produce products using the powers and the generative properties of AI. Excellent.

JP: And one of the phrases that you guys use a bunch when I hear you talk behind the scenes,

[02:17] What hidden knowledge looks like inside a company

JP: it's the phrase hidden knowledge. Yes. What is hidden knowledge and what's it look like inside a company? and where everyone holds a piece of it, but nobody has the whole picture. Absolutely.

CB: So the head knowledge is a very tricky problem in companies. So what happens with when you're working in an enterprise is you deal with a problem of opakness inside of a company where you are a grunt worker or you're a project manager and you don't know what's on the mind of the CEO. The one project manager doesn't know what another project manager knows. One developer doesn't know what the physicist knows in the company. And so there is a whole bunch of disconnect that happens. And all this lack of connection that is happening, we like to term that as hidden knowledge because it's just knowledge that's there. It's present, but it's hidden from all of the individuals who need to interact with it. And hidden knowledge actually has a whole bunch of nuance to it as well. There could even be hidden knowledge. You get a fact or you get a piece of information, a report, or anything. And you have this nugget of information, but the understanding of that information as well, how that information relates to itself, how that information relates to other things, that is also a form of hidden knowledge that kind of sneaks in there that you don't realize you're battling against when you're inside these companies.

JP: How much of this hidden knowledge is created by the culture of companies, Chris?

[04:04] Culture, job security and “CEO flybys”

JP: Because I think for most people who work in a corporate structure, they try to have as much, offer as much value to a company as possible. So it's the, you know, I want to be, I want to make myself irreplaceable, right? So there's this idea of like, I don't want to share all, all this hidden knowledge that, you guys are talking about. I don't want to share that because if they take my hidden knowledge away, I won't have a job. So how do you foster a company? And I know you're not the culture guy, but like you're trying to connect what's happening because of what companies have created through culture in terms of people, you know, CYA. Cover your ass, right? So tell me how this works with trying to keep that hidden knowledge out and everyone else be having access to it.

CB: So that is actually a very interesting question, and that starts to tickle on the idea of the user experience of Project Zero, because this is a user experience problem, because you're dealing with, like you've said, you have hidden knowledge that is cultural and embedded inside the company in ways that people don't really think about. One would be like what you just said, where they're going, this is job security. And then you also have cultural hidden knowledge that comes about from what a lot of companies are familiar with CEO flybys, where you have the CEO come in and he suddenly launches projects on all these people unsuspecting victims inside their company. So the and but those actions and those and what they do is actually generating a hidden knowledge problem because suddenly he spawns this little thing of knowledge over here and then it has to propagate around and it gets all mixed up. and muveled with the telephone effect and all that fun stuff. So what that means for Project Zero is a user experience problem. We have this platform where you're able to get together and you're able to start to share and aggregate and connect all that knowledge with each other.

[05:59] Sharing knowledge without breaking security

CB: But there is still necessary processes inside companies where you have security concerns where you shouldn't be sharing everything, right? But there is this ability that AI, is beginning to unlock for this user experience layer with all of this, that now hidden knowledge will be able to go into privatized containers, but also summarize and expose what is needed externally outside of that. So you no longer become a black box container where people are going, hey, who do I talk to? Where do I go for this? And like a month later, we found three contacts. And these three contacts are going, you need approval from blah, blah, blah, blah, blah, blah. You know, it is a nightmare inside companies that is a bureaucratic slowdown that happens with that. But with, when you do it properly inside a knowledge platform, you now have the ability to hide knowledge that is needed to be hidden, but expose what is needed to be exposed via summarizations, via filtrations of how information comes out. You can have all your knowledge base, but then expose certain nodules of information that needs to go out to others. So you overall start to have a user experience that is now able to offer more information than you would have been able to before, but it still is able to work within the security confines of the company. And then, like you said, you now also have a cultural problem of like all people job security and all that kind of stuff but the user experience also starts to draw people in

[07:49] Getting credit for your work

CB: into the experience of realizing you're contributing to the larger picture and you get and as project zero is being developed and is at its forefront being realized is that it is about a user seeing themselves inside the product so now they are able to see their impact they're able to see how they contribute and it becomes a good and useful experience because now counter to your example of job security, you now have the inherent ability of attribution. We all know in companies, attribution is a huge issue where you just go, hey, my managers took my project for me. You go and he took the credit and what the heck meant. And so now we have this ability. Now attribution flows. So there's a lot of also user experience, good feedback user experience that gets them into this idea of being able to work together and collaborate with their knowledge. All right.

JP: So, Chris, I want to pull on that thread about slowing things down within companies. And in the theme of slowing things down, what do you think is standing in the way of AI

[08:57] What’s slowing down AI adoption

JP: adoption inside of most companies? What's slowing it down there? Like, what do you think is how? Where is the pain point that's stopping it from really fulfilling what everyone kind of sees as the future, but it's going to, it seems like it's taking longer than people expected.

CB: So a lot of that slowdown comes from a lack of understanding of the tools that people are utilizing. There is always a basic knowledge of when you see something new and shiny that you get into it and you go, oh, this is what was provided. This is, goes back to that cultural talk that you're discussing, where especially in larger companies, they are given tools and they buy SaaS solutions. They buy all this kind of stuff and they expect a tool to work in a certain kind of way, but they don't see beyond the SaaS solution. They just see it as this is something that gets something done for me. So a lot of developers were, hey, we were given cloud code. And so a lot of developers went out there and they're like, I could code faster. And so they didn't see the bigger picture what was happening. They didn't understand that these AI systems are truly becoming more and more intelligent. They have this ability to communicate with each other and they have this ability to offer a freedom of information flow through the whole system, especially the enterprise. And that right there is probably the bigger win for AI that people don't grasp on to is that it's able to connect people. It's able to connect ideas and it's able to connect progress. It's able to connect your ticketing systems. It's able to connect everything together. in a faster methodology than people ever anticipated. The people who have started making those connections, where they started connecting these systems with AI, is able to communicate with each other, and they have good user experience of how they interact and oversee those AIs. These are the systems that are seeing the explosive growth and the, oh, I get it moments, where you're going, wow, this is, this is changing everything. It's even, I would dare say, probably more in startup still right now, but I would say the people who are making these connections and who are making similar kind of technologies along the lines of Project Zero are having those aha moments and are seeing this changes the organization. This changes how we interact with each other. This change, this changes business plans. You know, it's opening up everything to newer ideas. that we didn't really anticipate before, which is the goal of something as revolutionary as AI is.

JP: How much of a problem do you think is that old school entrepreneurs and CEOs

[11:44] The “credit card AI” mindset of legacy companies

JP: still transactionalize AI? Like, okay, my board says I need to get more AI, so I'm going to have my tech guy put it on a credit card, sign up for an LLM, we'll incorporate it in the company. Oh, okay, we want to just, we should just build it on. All right, it's cool. I'll just write a check and we'll build that, blah, blah, blah. And then so, and it seems like that old school mentality doesn't embrace the mentality you've just broken down there, this idea of like, it's more than just, here's my credit card, give me AI. How much of a problem do you think that is right now currently? Because I see that a lot with more legacy companies where they're like, I guess we got to do this. So we expect this ROI if I spend this amount of money on AI. Tell me about that idea. because this is what I'm seeing from the outside.

CB: Yes, and this is actually, this is, that, that mentality was really strong starting about, about two years ago, two, three years ago. And it really grew company, into company cultures using all of that old way of thinking of, hey, I spend more, I'll get return on investment. And, of course, that gets back into these are, this is just people having a hard time

[12:59] The lesson from computers in the ’80s and ’90s

CB: comprehending the future of what is coming at them. This is like, this is the same issue as like when the, when computers started hitting in like 80s and 90s of where people were going, oh, wow, this is like cool and this is neat, but now I could do my typewriting on a screen now. And that's as far as they saw. But they didn't see that, oh, these computers connect to each other. These computers now automate things. These computers do all this. The same thing is happening with the mentality with AI and larger companies and business mentality with it where they're going, oh, hey, cool, I have this AI tool. And yay, it writes my documents for me. It does it. It's almost an exact repeat where you're just going, no, no, no, no, no, no. This is a, you need to say, you need to understand the technology first.

[13:59] AI as a “capable individual”: the house-building analogy

CB: Like, what is it doing for you? like what it's capable of. The one example I have of like someone starting to grasp and get the idea of like, I didn't think of AI doing that for me is just the simple idea of categorization. So if you take in AI and I was saying back in like 2021, I was telling them when like LLM first started hitting, I was like, guys, you could tell this LLM to come up with some options and this LLM could pick it. And so I was like, yeah, I just use an LLM to pick one of those options. And that alone was a concept still today that gets people to kind of have their brains explode because they go, oh, you mean it could choose things. It could actually, it's thinking. And then I started explaining to them going, yeah, this is an AI that's in its own weird, intelligent way that you could tell it to sort a list. And it's not doing it like a program. It's thinking it through and placing that list in order. And so it's starting to realize like what it means to have generated intelligence and making it think of like this is a very capable. I gave this example to my CEO a little bit ago where I said you could think of this as a capable individual. So if you gave a really smart guy who has never built a house, right? But he's really smart. But, and you gave him suddenly this manual of step-by- step instructions of how to build a house, this smart guy is going to be able to build the house because he's smart, you know. He's going to figure it out and he's going to be able to do it. The same thing is kind of happening with AI, except now we have a really capable AI, dude. And you just have to break down your, you have to be able to break down what the process is. You have to break down what is going on and actually communicate to it what you want. And it could accomplish that if you break it down enough, if you give it enough knowledge to be able to do such a thing. And so that is also one of the heart things with Project Zero is this concept that there is always a minimum amount of knowledge that could accomplish the goal that you want.

[16:19] Alignment: giving AI the minimum knowledge it needs

CB: And I know that this is a buzzword in a, and it's used for safety concerns in AI, but I tend to call that alignment. And this is human alignment. It's aligning with what I want it to do. And so that's another important part about Project Zero and this whole concept of knowledge management and making sure that we are understanding that this is where we are going with our business is with this idea of knowledge gathering and cultivating. and making that our system. In fact, that's what the business is turning into

[16:59] Knowledge graphs in plain terms

CB: is now we're cultivators of this knowledge and this alignment problem. And we're outputting that.

JP: And geniant is solving this alignment problem with knowledge and knowledge graphs. And I know you and I have spoken about this in the past about how you've kind of taken this old school terminology and brought it into the 21st century. But in plain terms, why does it matter for solving hidden knowledge? the knowledge graph.

CB: Okay, so knowledge graphs, yeah. It, like you said, it is an older term, and they tend to call it back in the day, they used to call them like semantic networks. And that was kind of where they got into a whole bunch of stuff about, like, how sentence and grammar worked and everything. And the interesting thing now is LLMs basically solve semantic networks for us. That's kind of where you encode your semantic network now is inside an LM. But now I'm kind of promoting it into a higher level thinking of where we're knowledge graphs. So knowledge graphs are where you're having larger concepts, larger ideas, larger thought out processes or facts or this kind of this and that. And you're able to put them into little packets of information. And then you have other packets of information that you start bundling on. And then you start connecting those packets of information with each other. Now, these connections and these packets, you usually call them nodes and edges and like all the really nerdy math guys. This is graph theory and all that kind of stuff. And I'm like, yeah, yeah, yeah. But the bigger thing about this is that it is now AI-enabled graphs. So how this graph is connected is the alignment that we generate with it. So we are now putting our information into this graph. and we're able to connect all of our information. We're able to connect our ideas. So we have trains of thought where you could go down your graph and connect two separate ideas and have a train of thought of just how these two ideas can be related possibly. And these graphs are how we feed in aligned knowledge and aligned concepts into the AI as the context for the AI to begin work. and then AI is able to produce based off of that.

JP: May I ask a silly question, Chris, like how does a knowledge graph look in the world of LOMs? I get the idea of how it looks and you broke it down in math. There's numbers and there's pie graphs and all things. How does this graph look like in the world of AI?

[19:39] What a knowledge graph looks like (the Wikipedia example)

CB: Right now, you could go down a whole bunch of different trains. geniant is obviously generating a proprietary type of knowledge graph that we are confident and that's going to be a good system for managing the whole large basis of information. But one simple way to look at a knowledge graph is, and everyone who's messing with AI is familiar with already is basically markdown documents that have links to each other. And you may be familiar with that as Wikipedia. So it's a, the Wikipedia is a giant knowledge graph. It is article with a whole bunch of information. That could be a node. And then it has links and citations to other articles. Those are edges over to other things that you'll want to, that are related to each other or how they interrelate and derive with each other. Now, an interesting thing about, like, Wikipedia and why that's a little weak is often those links. We'll just link over and you have to kind of get an idea of what the edge means. like why is it linked in there? You kind of have to infer that out of the article and stuff. And in a knowledge graph, you would be more inclined to actually put the reasoning why that link is there. Interesting. In the link itself. So it would be, yeah, so you would have a stronger correlations with each other. You would have stronger understanding with that. And these edges and, nodes, by the way, are we've been rooting it in philosophical concepts. Like, we're working with ontologies and methodologies and all, histonologies and all that kind of ridiculous verbiage, but it's kind of guiding the principles of how these, this knowledge is organized inside the graph. So those are interesting topics to kind of peruse and investigate when you're kind of exploring the ideas of knowledge. Excellent. Thank you for humor for me on that.

JP: I was curious because you always hear a lot of this talk about AI and how it works. But like, as a tactile human, I'm like, so how is it like, what do I see? Like, how am I seeing it? So you created a beautiful picture of that. So thank you very much. So how's it work and where does it start? But I want to start with, I want to break these down into something you mentioned earlier, user experience and developer experience. So tell me about those.

[22:14] User experience: the ticketing system as a knowledge graph

JP: different experiences, how it works where it starts?

CB: Project Zero is a multi-stepped project where we're starting at some basis modules and it's evolving into the greater vision as it goes, obviously. So at its core, and I've been really enjoying working on this part of the project where I'm trying to make it the first experience. So this is like, I'll kind of describe the first experience. the first experience, because it kind of helps people start to see and understand from a viewpoint that they already get. And that is, it's being built off of a ticketing system. So you have linear, you have Jira and all that kind of stuff. And everyone has worked in a ticketing system. We all hate them. But there's wild usefulness to it. It's a workspace for humans to work together in parallel. And a ticketing system, likewise, it works really great for agents. Agents could go in there and make tickets, sign tickets, and they work really well. It's a fantastic space. So the ticketing system is actually the first of the user experiences of a knowledge graphing. Why? Well, if you think about tickets, tickets are nodes of work and a bunch of information associated with it. Tickets have dependencies on other tickets. You can't do this ticket until this ticket is. done and so and also you are able to create cross-references with your tickets you're able to do a whole bunch of stuff with tickets as real similar to a knowledge graph and so this is like a very interactive knowledge graph because now humans are working alongside this ticketing system so what i did was i created my own ticketing system because why not we're in the age of AI so it's like just built my own so i built i built my own ticketing so we built our own ticketing system and inside this ticketing system it has the ability to have very strong distinctions between what is agent work and what is human work and so we you're able to make tickets be locked to humans so agents can't touch it or you throw it over to agents and anyone can touch it and then you kind of manage these tickets and you kind of get them

[24:29] Making a flood of AI output manageable: a visual graph of the work

CB: moving around so interestingly what is developed and what is generated into the user experience of AI and where it's been lacking and this is one of the first items that I was. wanted to tackle is AI is an overwhelming amount of information. So when you get, you do anything with AI, especially programming and stuff, it's going to send off these lourbs at you that take about 10 minutes a piece to read. And you, it's a lot to digest. It throws and stuff that you just do even think about on whatever. And sometimes it says things in cryptic ways that mean absolutely nothing because it's, you know, hallucinating out of it's mine. But the, so the, so I realize, well, what if I just stop thinking about all the information and it? It's just start thinking about the tickets. I just want to start working with the graph, so to say. And so I bubbled this up into just showing the graph, where now we have tickets that are just showing those little bubbles. And the bubbles are starting to show me this is human stuff and this is agent stuff. I can investigate as I'm going to investigate, but I have a large picture of my entire organization of all the projects that are running. I'm running about six to ten projects inside the ticket. system by myself right now. So literally just me going in there and investigating and doing stuff. And then you wire up the system with automated agents and watch it fly. And it's an incredible experience. So this is just kind of like a taste of the idea of we need to start like understanding that graph is there. We're working with nodes and all of that. And we need to understand it's a graph, but also not have to deal with the graph. Because unfortunately, while I've been explaining knowledge graphs, I'm trying to explain them to you and make them understandable and all of that, they are from a human's point of view, really hard to comprehend. It doesn't matter what you, it doesn't matter how smart you are. The greatest minds out there have been working on graphs and writing proofs for graphs. And just recently, AI has proven several of the great minds wrong on several. graph theory items. And so it's a that's how hard graphs are for us to comprehend. So I'm, we're working on abstracting the graph away while moving us into this working alongside the graph and alongside the AI while the AI manages all the details for us. So what this moves on into of what it can be.

[26:59] Generating reports, charts and tables on demand from your knowledge

CB: I just told you about ticketing and ticketing is doing work. So what you can imagine from that is it's generating stuff for you, right? So it's generating work. It's generating items for you. So that means the next step that happens out of this is you could have a generative experience with your knowledge. So what Project Zero eventually gets to and the point that it comes to is one, you have this system that you're working together in, but then your experience, with querying for it is something that could be completely generative. You have a ticket and you could say, make me a report off of this. Tell me what this is. Give me a graph. Give me a table. Give me a block. And so your user experience starts to explode that way. And because you're able to be interact with all of your information in ways that you didn't really expect to be able to interact with in a faster. means than you expected it to happen with. And suddenly, the user experience becomes, oh, I don't have to worry about this anymore. So what am I going to do with that? And what am I going to do next? And so your user experience then becomes what, what's the future? What is the next great thing that we can make now that this is all done?

[28:29] How hidden knowledge gets collected: Sense & Extract

JP: Okay, before we get to next, I want to talk about what we do now to get to next. And that is this hidden knowledge we talk about. Where are we getting it? How are we collecting it? Is it like, I'm going to dumbify. Are we all dumping things into a drop box and then Project Zero does the rest? How are we collecting this hidden knowledge? And then I want to get to what happens next. But how do we do that part? Okay.

CB: So that is actually, so Project Zero is divided up into several modules. And so on one end, we have this. problem that you're talking about right now. How do you get information? And you move through the brain of Project Zero and then you move to that futuristic view and generative experience. So the first two modules that I'll talk about right here is called, for us, Sense and Extract. So there are two different kinds of concepts of how you get information like as a human and also as an AI. And one is sensing things. You have to take in your senses. You have to go out there where there's things happening and feel it out and have it happen. So this includes things like integrating with your Slack channels. This includes things like integrating with any real-time feeds that's coming out of your system. This includes, you know, data lakes getting new reports and all that kind of stuff. So that is kind of a sensing platform where in real-time you're consuming data and it's coming inside of the queue. It gets analyzed and it gets broken apart. into relevant and meaningful information.

JP: So it acts like a sponge within a company in this project. Like it's whatever it's on that's happening, whether it's email, Slack, in the network, in people's drives. It's all being sponged up, right, by this part of Project Zero. Okay. Yes.

[30:19] Not Big Brother: organizational knowledge, not employee monitoring

CB: And this isn't a lot of people will hear this and they'll immediately go over to the big brother companies out there that want to, you know, monitor employees. But this is different because it's not monitoring for the sake of giving you a report on what's this employee doing it's giving it's sensing for the sake of understanding your organization so it's actually it's taking in the information for understanding the business problem it's trying to uh extract this information and go oh this there's new ideas being presented here we should remember these ideas we should remember we should expand and expound on these ideas do these ideas have overlap with our previous ideas you know it's more it's more focused again on organizational knowledge. It's not necessarily, I mean, it could be maybe, but in its initial form, it's not going to be really concerned with operational knowledge, right? Because it's not a, you're trying to get this to get the picture of your company and make your company productive with AI.

JP: It doesn't care what Jeff from accounting had for lunch today. It's not taking that information. Yeah, yeah, yeah.

CB: Or even how long you took a lunch, like, who cares? Sure. Yeah, because that is another thing that kind of makes an interesting point to all this, as a lot of this becomes kind of time invariant where you're less concerned about time. You're just more concerned about ideas. So that's the idea of sense. And then, of course, there's extract, which you could kind of hopefully extract what it's going to do. Sure. So extract is essentially the more manual. approach of where you have like legacy information or you have large pools of data or something that is just sitting there that has something interesting about it that you need to have analyzed and you need it to be kind of pulled in. So that's that's that drop box idea. You said, is there a drop box somewhere? Maybe that could be a part of extract. And so you just throw it in there and its whole system is meant for the uptake of that. And of course, that is like a system that you kind of tweak and tune, you do a little prompt engineering and all that kind of stuff to make sure that it's aligning itself correctly with how it's pulling the information out. And of course, the one since an extract or pulling in the information and coming

[32:44] Humans in the loop: experts signing off

CB: into the graph, there is the user experience of the graph and how that's cultivated. And so that information, of course, is cultivated through humans and aligned through humans where you still have managers and you still have experts who are signing off on. going like, yeah, this is relevant to our problem or, hey, this is true, but we don't care, you know, and, you know, so there's still, there's still like human involvement. Sure. Because again, it's human alignment. It's, uh, we need it to keep understanding us and not doing whatever it wants with this information.

JP: Excellent. All right. So once an organization has done those three steps, everything centralized and map, what becomes possible that wasn't before this process?

[33:29] What becomes possible: software factories and beyond

JP: Give me like one example. And I'm sure, Chris, knowing you, you probably have 75 examples. Give me one example of something that would be possible after they've gone through this process of working with Project Zero. Okay.

CB: So a possibility. There's obviously the easy example to give, which you say, it's a software factory, which everyone in their dogs doing right now, because that's the natural course of things. We saw it when developers are developing it. And so we love software factories. And so that's a natural output of such a system. And it would be easy to do because now your software system, you extract the knowledge of the, like, the architecture of what you would be doing, right? And so that's part of the knowledge tree. And then you extract the business problem. And when those two meet, that is what produces. That's the software factory. It produces the software. So this, though, enables such abilities that now the marketing guy has your architect's knowledge in the system.

[34:44] Marketing, CEOs and work across departments

CB: And so the marketing guy can also make a system. He could produce a product for marketing. That is business aligned, architecture aligned. it is able to produce something completely unique to marketing. That marketing was never able to produce before. So marketing is suddenly given this amazing ability to work beyond their normal means. They're able to produce something brand new that neither you or I have probably even thought of. Because since Wynn has, I mean, it's rare that you get cross-discipline idea people who have moved across disciplines and they put them all together and come up with this cool idea, right? Well, now we're enabling more people to work into that cross-discipline zone. And so now they're able to start producing things that we haven't imagined before. So the, and this includes your CEOs and stuff. Now CEOs are no long, they don't have to talk to your engineering department and going, I really want this dashboard. For some reason, CEO's always going for dashboards. And so he, but now finally, the CEO's getting his dashboard. but he doesn't have to harass engineering who has other kinds of problems to solve and he has this enablement to get this visibility into this company and all that kind of stuff and I know a lot of these are examples are kind of contrived in the sense of oh but this is kind of like stuff that we did with people before right but where it gets fascinating to me is that it it opens up the human capital aspect of it, where I've been saying that, oh, but now you have all these people with free minds and these people who are now intelligent, and you know that because they work in your business. So you know you have a smart group of people who now are working to identify brand new things to work on, and they're able to also have disability across the company. so they are able to do things that benefit the company in huge ways that you just, again, I say it a lot now, because I keep telling people, I'm like, it's going to bring out new stuff. You don't know what it is yet. I'm like, it's a, it's brand new mixtures, it's brand new freedom, it's brand new ideas and exploration that we just haven't had. And it's, it's nigh infinite possibilities because it's so much connection. There's so much ability to see across the vistas of the information in your business. That's great.

[37:29] The hard part: “finding the lie”

JP: That's great. And like many AI guys, Chris, you make everything sound easy. Like, you know, all these AI guys are like, we're going to solve for cancer and for poverty and all this stuff. And not saying that you're saying that. But the reason I bring it up is what's harder? What's the hard part of aligning hidden knowledge? Because that's the thing. Like, if, you know, if someone listened to this conversation and we're watching and be like, oh, okay, cool. this all makes sense to me. And of course, you, very positive, very optimistic about it, making it sound easy. But what's, what's, give me a little, give me a little something gritty here. What's going to be hard for people to do when it comes to aligning this hidden knowledge?

CB: Finding the lie. So if you are, um, humans are gullible. So we, uh, we tend to trust something that sounds authoritative, right? So the, um, in a system when you're building, a system like this. And you're working with LLMs and AI models. We all know about hallucinations thoroughly, right? And so we know they lie. And so the, what this again gets into why I am saying the biggest thing that Project Zero is digging into compared to most other projects that are approaching this. It's a U.S. problem. It is very much a user experience. problem, and that is a very hard thing to solve because you fall into complacency with AI.

[38:59] Don’t surrender your abilities to AI

CB: You get into a mode of what I'm calling, what I have been telling a lot of people in my company, I'm saying, I am working really hard to make sure that humans don't surrender their abilities to AI. We're not sitting here trying to surrender everything to AI. We're trying to work next to it. But the tendency for us, because we're lazy people, and engineers especially, we're really lazy, is that we love surrendering things to things. And so the user experience with this, you have to keep them, you have to keep the person involved. They have to feel the contribution. They have to feel like they themselves are in the system in a way that's benefiting them and it's benefiting others. Because when you do that, you're trying to create this nice little feedback loop with your user of going, hey, this is good, and I want to make sure it stays good. I don't want it. I don't want it to fall apart. And the, of course, there's like all these safeties and stuff, which is another hard part about the whole system. You build in safeties.

[40:09] Keeping the genie in the bottle

CB: You build in a ton of safeties. Like, you're just, you're trying to isolate that AI. You're going like, no, stay off my production. You know, you're trying to like bottle it up and put the genie as much on the bottle as you And that's really hard. That's another hard part about all this is keeping that genie in the bottle. But letting out just enough of smoke to still be useful, you know. So that I would say is the biggest challenge, but also one of the biggest, I feel like going to be the biggest successes of Project Zero is because we are working really hard on that user experience loop to make sure that the users are truly expert. working together with the AI systems. Excellent.

JP: And lastly, Chris, if this is the first time folks are hearing about, seeing about Project Zero,

[40:59] The one takeaway

JP: after, you know, checking us out here today, what do you want them to walk away with? Like, what do you want them to feel like they learned? I mean, obviously a lot. So, you know, some notes were taken and some transcripts will be, you know, pumped into an AI after this. But like, what's one takeaway human can really walk away with after our conversation?

CB: The big takeaway about something like Project Zero is one, not doubting that something like this is totally possible right this moment. And then two, understanding how amazing and good it can be. Because again, this is one of the biggest user experience problems of the whole thing is understanding how hopeful. and great such a product is and that it is what is coming. It is what is happening and it's already happening in some companies to some degree or another. Something like Project Zero presents a whole new way of producing that we have never had before, which is going to open the doors for all of us to have a brand new outlook on what on work and knowledge work as a whole. And I think we are all going to accelerate into something grander than when we're at right now, obviously, just like the computer coming along. And we're going to see how much a project like Project Zero will just handle immense amounts of work for us that we just still have a hard time grasping. and enable us to experience something newer than we also are having part-time imagining. Good stuff, Chris.

[42:59] Coming up next: knowledge acquisition in practice

JP: And so for our next episode, what should we cover that we have not covered so far in Project Zero? What do you think should be next on the docket?

CB: I think we're going to need to look over some of the ideas of more of the grounded reality of what it takes to make Project Zero happen. And this will get into a lot of the knowledge acquisition and what it looks like today with how we send researchers into companies for the sake of essentially generating knowledge graphs right now. It's a lot more of a manual experience, obviously, because we're using researchers. But that gets into a lot of the really cool nitty-gritty of just how important knowledge is, how what knowledge does for the sake of building a project and what knowledge does for the sake of making sure you get accelerated development out of AI that

[43:54] Wrap-up

CB: aligns with your intent and your business practice.

JP: That's wrap on episode one. If your team has an AI pilot that worked but hasn't scaled, that's exactly the problem geniant solves. Go to geniant.com to learn more about Project Zero and start a conversation with their team. Subscribe so you don't miss the next episode and share this one with a leader on your team who's trying to make AI work at scale. I'm Joe Pardavila. See you next time on The Road to Project Zero.