Latest / The Adults In The AI Room / Earley AI Podcast – Episode 90: Federated AI, Decision Intelligence, and the Data Architecture Reset

Transcript

Earley AI Podcast – Episode 90: Federated AI, Decision Intelligence, and the Data Architecture Reset · The Adults In The AI Room

• 0:21
  SPEAKER_01: So, Christian, are you just uh okay going all right?
• 0:24
  SPEAKER_02: Yeah, I'm gonna I'm gonna go dark here unless you're gonna be here.
• 0:27
  SPEAKER_01: Okay, that sounds good.
• 0:28
  SPEAKER_01: Uh Todd, nice to meet you.
• 0:31
  SPEAKER_01: Likewise, how are you today?
• 0:33
  SPEAKER_01: Good.
• 0:33
  SPEAKER_01: Where are you based?
• 0:35
  SPEAKER_01: I'm in North Carolina.
• 0:37
  SPEAKER_01: Okay, I'm out of the Boston area.
• 0:39
  SPEAKER_01: Okay.
• 0:39
  SPEAKER_01: Um was having some computer challenges this morning, so I was a minute late, but uh I think I think I should be all set.
• 0:46
  SPEAKER_01: And let me try to get my colleague out of the room here.
• 0:51
  SPEAKER_01: So Todd, uh, how did uh you come across our radar or we come across your radar?
• 0:56
  SPEAKER_01: Tell me, tell me uh how we got connected.
• 0:58
  SPEAKER_00: Um Kristen uh from Grid House, our uh PR agency found you, I believe.
• 1:05
  SPEAKER_00: Um Kristen, you can fill in the blanks there.
• 1:08
  SPEAKER_02: Yes.
• 1:09
  SPEAKER_02: So just reached out, saw some of the things that you talk about, and thought you might be interested in connecting with Todd.
• 1:15
  SPEAKER_02: And I sent a note kind of on some of the bullets that we were talking about at the time.
• 1:19
  SPEAKER_02: And Carolyn thought you you guys might have a nice conversation here.
• 1:24
  SPEAKER_01: Sure.
• 1:24
  SPEAKER_01: And and forgive me, I I I don't have the bullets in front of me.
• 1:28
  SPEAKER_01: So Todd, why don't you give me a kind of the lay of the world, the lay of the land, the world according to Todd, and uh we'll talk about how this might fit in with uh uh what our audience is interested in.
• 1:39
  SPEAKER_01: So give me a short ground and what you're doing and your kind of philosophy on AI.
• 1:44
  SPEAKER_01: I know that's a lot, a lot in there.
• 1:46
  SPEAKER_01: Broadbrush strokes is fine.
• 1:47
  SPEAKER_00: Yeah.
• 1:48
  SPEAKER_00: Um broadbush strokes.
• 1:50
  SPEAKER_00: So uh I'm CEO of a company called Axonis.
• 1:54
  SPEAKER_00: Um I uh feel like this was a bit of a diamond in the rough from the perspective of uh my background is um just kind of high growth enterprise software, often open standards, open source, often developers related to.
• 2:12
  SPEAKER_00: Uh and so uh I've I was introduced to this technology that was incubated inside of a government solutions integrator.
• 2:23
  SPEAKER_00: Um and we spun it out in December.
• 2:27
  SPEAKER_00: And the technology was an AI execution platform with a unique um a unique element to it, which is a federated data architecture.
• 2:40
  SPEAKER_00: Essentially, what that means is we send models out to data as opposed to having to centralize data.
• 2:46
  SPEAKER_00: Um, we can train models on data in this federated environment.
• 2:51
  SPEAKER_00: That's called federated learning.
• 2:52
  SPEAKER_00: You may have heard of that.
• 2:54
  SPEAKER_00: We can also serve models out to that environment so that inference happens where the data is being created.
• 3:01
  SPEAKER_00: Um, what we also realized from that is that where we play in an AI architecture gives us the unique opportunity to capture artifacts around decisions that are being made with AI.
• 3:18
  SPEAKER_00: And so we have a layer that captures the evidence from the data that you're using, the model, the chat transcript, if you're using a chat, or if you're an agent, it captures the agent.
• 3:32
  SPEAKER_00: Um, and all the way to the human in the loop decision.
• 3:36
  SPEAKER_00: And that now becomes an artifact in the system.
• 3:39
  SPEAKER_00: And once you have decisions as data, you can then train models on those decisions and continually optimize them.
• 3:46
  SPEAKER_00: So you kind of have this flywheel effect.
• 3:49
  SPEAKER_00: So, and the only way that that really becomes useful is if you have this federated access to the data.
• 3:55
  SPEAKER_00: And so we kind of put put that all together.
• 3:58
  SPEAKER_00: If you think about for regulated industries, if you think about regulated or hard-to-reach data um that's resistant to centralization, um and centralized data sources can just be a source for us as well, you have this pretty powerful system for uh you know high consequence decision workflows that uh you know you're gonna see as custom companies start to transform their businesses around either agents or just AI in general.
• 4:33
  SPEAKER_01: Okay.
• 4:33
  SPEAKER_01: And then go ahead, keep going, go ahead.
• 4:37
  SPEAKER_00: On top of that, we realized, you know what, we shouldn't just have this.
• 4:42
  SPEAKER_00: Everybody should have this.
• 4:43
  SPEAKER_00: And so we are working towards an open standard around this decision evidence specification, where if AI is being used by agents or people, this should be an open standard where we define the data by which these decisions are captured for auditing, for regulators, for improvement across across industries.
• 5:03
  SPEAKER_00: And so we're working with a few organizations to try to make that decision evidence specification happen.
• 5:09
  SPEAKER_01: Okay.
• 5:09
  SPEAKER_01: And and uh just to take a step back, uh, how long has the company been around?
• 5:14
  SPEAKER_01: And um, what was your background before how long have you been with them?
• 5:18
  SPEAKER_01: What was your journey uh up till this point?
• 5:20
  SPEAKER_01: A little bit of more context.
• 5:22
  SPEAKER_00: Yeah, the technology has been around for the last two years inside of this other company.
• 5:27
  SPEAKER_01: Okay.
• 5:28
  SPEAKER_00: Uh and then we spend who is the large, who is the integrator that um they're called T2S, and they're an integrator for the Department of Defense and the intelligence community.
• 5:36
  SPEAKER_01: Okay.
• 5:37
  SPEAKER_00: So they they operate in environments that are uh where the data is top secret often.
• 5:44
  SPEAKER_00: And so they have you know the data security uh technology that we use is you know classifies data down to the cell level if you need to.
• 5:54
  SPEAKER_00: Uh, and then they operate in uh DDL environments, which might be distributed, disconnected, you know, air gapped.
• 6:01
  SPEAKER_00: And so the technology kind of runs in that environment.
• 6:04
  SPEAKER_00: So again, it has kind of all the features you'd need for regulated edge industry um from a data uh protection perspective as well as from sort of an operating perspective.
• 6:16
  unknown: Okay.
• 6:17
  SPEAKER_00: Um my background, yeah.
• 6:19
  SPEAKER_00: So I've I've been uh a part of multiple infrastructure transformations.
• 6:27
  SPEAKER_00: First one being Unix to Linux.
• 6:29
  SPEAKER_00: I was at Red Hat in the early days uh and kind of saw how the infrastructure changed from you know proprietary to high performance open standards using Intel architecture.
• 6:42
  SPEAKER_00: So that was number one.
• 6:43
  SPEAKER_00: The next one was DevOps.
• 6:45
  SPEAKER_00: I was part of um a company called Ansible that got acquired by Red Hat.
• 6:50
  SPEAKER_00: And this was um, you know, infrastructure as code.
• 6:54
  SPEAKER_00: And then I was part of kind of DevOps as part of GitLab, where we kind of helped move you know the industry from a whole bunch of tools to a more standardized group of tools for uh you know DevOps and development.
• 7:11
  SPEAKER_00: And then the last one, I was part of um Chainlink Labs in the blockchain space.
• 7:17
  SPEAKER_00: We were building infrastructure to move banking from you know its current system to tokenization, which is now really taking off.
• 7:25
  SPEAKER_00: Um, so that's my background.
• 7:26
  SPEAKER_00: It's just infrastructure transformation, which is why I'm here.
• 7:29
  SPEAKER_00: I thank God there's gonna be a lot of enterprise infrastructure transformation happening.
• 7:34
  SPEAKER_01: Yeah, certainly.
• 7:36
  SPEAKER_01: That makes a lot of sense.
• 7:37
  SPEAKER_01: And you and your role again in this organization?
• 7:39
  SPEAKER_01: I'm a CEO CEO.
• 7:41
  SPEAKER_01: Okay.
• 7:42
  SPEAKER_01: And um, one of the things that we like to start our sessions, and we have a and thank you for that background, that was helpful.
• 7:49
  SPEAKER_01: Um one of the things that we like to uh begin, we have a mix of of uh audience uh types.
• 7:55
  SPEAKER_01: So we have uh people who are practitioners, people who are on the technical side, people who are executives.
• 8:00
  SPEAKER_01: And one of the things we like to do is, you know, kind of start with a problem statement or a misconception.
• 8:07
  SPEAKER_01: You know, what is it that organizations are kind of missing today from your perspective about AI?
• 8:13
  SPEAKER_01: What are the things, the uh misconceptions or the uh false premises or whatever it is that you wrestle with?
• 8:20
  SPEAKER_01: What do you where do you have to educate executives?
• 8:22
  SPEAKER_01: What are the things they're getting wrong or that um you need to disabuse them of, right?
• 8:28
  SPEAKER_01: Um, used to be okay, you know, you you throw all the you throw your your all you all the data at the AI and see what see what happens, right?
• 8:35
  SPEAKER_01: And people are kind of aware that you know you can't do that, you need reference architectures, you need to contextualize things, you need to look at the content and the data itself.
• 8:43
  SPEAKER_01: From your perspective, what are the things, what are the big misses?
• 8:46
  SPEAKER_01: What are the big things that you're finding that uh leadership and executives don't uh understand or misinterpret?
• 8:53
  SPEAKER_00: I think the big one is they mostly thought that they had made the investments to centralize their data and that AI would be automatic on top of that.
• 9:03
  SPEAKER_00: And there's two problems with that.
• 9:04
  SPEAKER_00: One is they never finish that project, yeah, and they never do.
• 9:08
  SPEAKER_00: So they always have data outside of that environment.
• 9:11
  SPEAKER_01: It's it's it's it's a moving target, right?
• 9:13
  SPEAKER_01: It's a it's an evolving ecosystem, so it never changes, never, never stabilizes.
• 9:17
  SPEAKER_01: Yep.
• 9:18
  SPEAKER_00: The second problem with it is it was built for BI, not AI.
• 9:22
  SPEAKER_00: So the problem space is that oh, I needed fast reports, I needed really clean data, I needed, you know, because the output was these real-time reports, right?
• 9:35
  SPEAKER_00: And then a human would analyze it.
• 9:37
  SPEAKER_00: And AI requires different things.
• 9:39
  SPEAKER_00: You know, AI, you want to inference like to really transform your business, you want to model inferencing on your dirty data as it's being created at the edge where the transaction is happening, whatever that whatever your version of a transaction is.
• 9:53
  SPEAKER_00: And that's why centralized data that's been cleaned and overly transformed, you know, is never going to get you business transformation using AI.
• 10:04
  SPEAKER_00: It's gonna start and you're gonna get some better analysis, but you're not gonna get agentic aid, you know, agents doing really fast things at transaction time to transform your business.
• 10:14
  SPEAKER_00: That's to me the big one.
• 10:17
  SPEAKER_01: So it's kind of the fallacy of you know, centralized data or you know, having a data uh warehouse or data uh or data fab, or whatever you want to call it.
• 10:28
  SPEAKER_01: It's you're not you're not centralizing, but you are using this distributed.
• 10:31
  SPEAKER_01: And I think a lot of organizations are kind of coming to that understanding that the data is going to live where it lives, where it's gonna be created in these places.
• 10:38
  SPEAKER_01: There's you know, it's it's a constantly changing ecosystem, right?
• 10:42
  SPEAKER_01: Everybody's operating on data, they're changing it as it's happening, they're transforming it.
• 10:47
  SPEAKER_01: Uh so so what is it?
• 10:49
  SPEAKER_01: So if you kind of lay that out as your premise, and again, kind of want to uh a punchy headline on that.
• 10:56
  SPEAKER_01: What is the what is the problem?
• 10:58
  SPEAKER_01: Then what are the some of the things that people need to do differently?
• 11:01
  SPEAKER_01: So so um so lay out that that problem or that that misconception, and then have some bullet points about what uh they're doing.
• 11:09
  SPEAKER_01: And and then and so you could talk a little bit more about that, or we could talk uh at for our purposes about you know what is the specific problem that you are solving.
• 11:19
  SPEAKER_01: So when you go into an organization, what is a what does a customer look like and what does a problem look like that you are trying to uh help an organization with?
• 11:29
  SPEAKER_00: Yeah, well, one thing I've I've found in in data is that, or in sorry, in AI space is that the the people who are driving the transformation are business people.
• 11:41
  SPEAKER_00: And then they're asking, you know, IT to help them, right?
• 11:45
  SPEAKER_00: And um so you know, when we originally started this, we were um we're a platform.
• 11:52
  SPEAKER_00: And so we're like, look how great federated data access is.
• 11:55
  SPEAKER_00: And and the IT people were like, that's amazing.
• 11:58
  SPEAKER_00: I I see what you're doing there, and it it would really be great to have this system because we know we haven't centralized all our data and we know we haven't, you know, we know all these problems.
• 12:08
  SPEAKER_00: However, um the problem that the business people are trying to figure out is business transformation, and that could be different for different folks, right?
• 12:17
  SPEAKER_00: Um and for us, the problem we solve is you're trying to integrate AI into high consequence decision workflows.
• 12:27
  SPEAKER_00: You need to have that be auditable.
• 12:30
  SPEAKER_00: You need to probably have a human in the loop at least for a while.
• 12:33
  SPEAKER_00: If not, you're gonna delegate agents with human authority.
• 12:36
  SPEAKER_00: And you are trying to make those processes much faster and much more efficient.
• 12:43
  SPEAKER_00: That's the problem that we solve.
• 12:45
  SPEAKER_00: The way we solve it is through this federated data access because they what what they get is a bunch of people coming in and showing them a WYSI demo that they created in Claude.
• 12:56
  SPEAKER_00: And then the gap between that demo and the ability to get that into production is huge.
• 13:03
  SPEAKER_00: It's enormous.
• 13:04
  SPEAKER_00: We're actually doing a disservice to a lot of executives by showing that stuff.
• 13:10
  SPEAKER_01: Say it one more time.
• 13:11
  SPEAKER_01: What was the last thing you said?
• 13:12
  SPEAKER_00: We're doing a disservice to executives who aren't technical and showing them a claw demo that looks real when the issues are around auditability, sure, access to data securely, um, you know, all the all the ilities that you need to make something actually work.
• 13:31
  SPEAKER_00: I I heard uh I heard a hilarious story of there's a New York Times article about a doctor, a radiologist who had his own practice, and he was uh their story was about you know how people are innovating at the edge with AI now because of Claude and Claude Code and all this stuff.
• 13:48
  SPEAKER_00: And the doctor had made middleware to connect some of his systems together.
• 13:53
  SPEAKER_00: And the practitioners who know what they're doing were like, there's at least 18 HIPAA violations happening right there.
• 13:58
  SPEAKER_00: And the doctor's quote was, you know, I love doing medicine, but doing engineering is thrilling.
• 14:07
  SPEAKER_00: And we were like, what if it were the other way around?
• 14:10
  SPEAKER_00: Where, you know, the engineer was like, I love doing engineering, but when I do surgery using Claude, it's thrilling.
• 14:17
  SPEAKER_00: You know, it's like it would blow your mind, right?
• 14:20
  SPEAKER_00: And so I think that's the thing is we're we're doing demo where it's really cool, a lot of stuff works, uh, and we are missing out on the reality uh that needs to happen from an infrastructure perspective to make all these things come true, right?
• 14:36
  SPEAKER_01: The um sorry, I'm getting a weird message up.
• 14:41
  SPEAKER_01: My computer's been misbehaving, and I'm wondering if I have a problem because it's been throwing weird messages and ask Claude.
• 14:50
  SPEAKER_01: Yeah, I will ask Claude, but it's like, what is distiller and why is it doing what it's doing?
• 14:56
  SPEAKER_01: Why is it asking me this stuff?
• 14:58
  SPEAKER_01: Anyway, uh all right, sorry about that.
• 15:00
  SPEAKER_01: So um, yeah, no, that makes sense.
• 15:02
  SPEAKER_01: And that is, you know, I've done a lot of research and writing on the pilot to production gap, right?
• 15:07
  SPEAKER_01: Lots of easy pilots, but then getting to production all the things that you just said, you know, you don't have the same data quality at uh when you're trying to operationalize, uh, you don't have uh it lovingly, painstakingly, you know, artisanally uh crafted, right?
• 15:24
  SPEAKER_01: You have yeah, you have these demos that, yeah, well, I'll take this data and I'll fix it and I'll enrich it and I'll do this stuff, and then I'll have this nice, wonderful demo.
• 15:32
  SPEAKER_01: And then getting that to you know, beyond the demo is very challenging and difficult.
• 15:37
  SPEAKER_01: So um what uh give me a sense of the customers that you have today, and that you know, how large is the organization?
• 15:45
  SPEAKER_01: And give me a sense of who you're working with and kind of maybe a story of of you know, kind of a before and after.
• 15:53
  SPEAKER_01: This can't be a sales pitch for your company, but it's always good to hear, you know, stories of what the problem was and then how a problem was solved.
• 16:01
  SPEAKER_01: And you know, obviously, you know, you can refer to your technology, but you want to you don't want to say, you know, um, you know, we did this, we did that, uh, you know, uh how do I pronounce the name of the company?
• 16:13
  SPEAKER_01: Is it Axonus?
• 16:14
  SPEAKER_00: Axonis, yeah.
• 16:16
  SPEAKER_01: And the Axonus tool did blah blah blah, and with Axonis, right?
• 16:19
  SPEAKER_01: You want to be careful about that.
• 16:21
  SPEAKER_01: But give us a sense of what kind of a uh uh situation that you walked into, how they had tried to solve the problem, and then uh how you were able to facilitate that if you if you could give us a story like that.
• 16:34
  SPEAKER_01: And then again, uh how many customers do you have?
• 16:37
  SPEAKER_01: How large are they?
• 16:37
  SPEAKER_01: What what is a what does that world look like?
• 16:40
  SPEAKER_00: Yeah.
• 16:41
  SPEAKER_00: Um, well, for your purposes, this isn't public, but we you know we just spun out in December.
• 16:46
  SPEAKER_00: So we have a handful of customers and a bunch of POCs that we're running right now.
• 16:51
  SPEAKER_00: Um, so I'll give you a couple of use case um thoughts.
• 16:56
  SPEAKER_00: So there's a company, uh well, you may remember uh in Texas last year, they had severe flooding, um, loss of life, um, some tragic uh situations.
• 17:10
  SPEAKER_00: That is required that has caused Texas uh municipalities to say, how do we get better at predicting these floods and make better decisions about when we um tell the public about when they need to leave or what needs to happen.
• 17:26
  SPEAKER_00: And uh there's a gap there.
• 17:28
  SPEAKER_00: So sensors exist, governments exist, people who have to make decisions, but the intelligence in the middle doesn't exist very well right now.
• 17:36
  SPEAKER_00: You can get data, but you aren't getting intelligence.
• 17:39
  SPEAKER_00: That's the gap that Exonus solves.
• 17:42
  SPEAKER_00: So we uh we're working with the company that installs and manages sensors, flood sensors in Houston and Galveston.
• 17:52
  SPEAKER_00: And uh we have built them a platform in the middle using our technology basically that uh allows um government officials to uh use AI in really smart ways to help them make decisions faster.
• 18:08
  SPEAKER_00: It's not just uh put AI on the problem, it's you got to get all the data in.
• 18:14
  SPEAKER_00: The data comes from disconnected, sometimes uh you know, decentralized systems.
• 18:20
  SPEAKER_00: You then have to have smart ways to visualize that data, put it in the vocabulary that they can see.
• 18:26
  SPEAKER_00: Um there's models in there that aren't just LLMs, there's there's ML models for prediction.
• 18:32
  SPEAKER_00: And then you have to make it really easy for them to walk through a decision with evidence from the data, create that decision artifact because there's legal liability around that, and then be able to spin off an action from that, and that's what we're we're helping them with.
• 18:51
  SPEAKER_00: Um, and so uh and do to do that in a much faster timeline with evidence.
• 18:57
  SPEAKER_01: Gotcha.
• 18:58
  SPEAKER_01: And it sounds to me like there's a um a governance and um traceability and auditability um angle to this, right?
• 19:07
  SPEAKER_01: Because you need to understand um what data sources you're using, uh, the condition of those data sources, the um provenance of that data, the rights that you have for that data, what the downstream applications can be, and the dependencies, right?
• 19:23
  SPEAKER_01: If something is changed upstream, what does it impact uh downstream?
• 19:28
  SPEAKER_01: And so would you say that your platform is um has governance um capabilities, or does it fit into a governance uh um uh infrastructure, or or how does that uh how do you operationalize that, or how do you take those decisions that are that you're saying are documented with the data and bring that into a decision-making system, decision-making bodies, decision-making rules, decision-making processes, right?
• 19:57
  SPEAKER_01: How does that fit in or work?
• 19:59
  SPEAKER_00: You know, governance can happen to different layers.
• 20:01
  SPEAKER_00: Um, there's model governance, our our system has model governance because we can host the models if you want.
• 20:09
  SPEAKER_00: Um, but I think for this, what you're talking about is um, I'm a decision maker.
• 20:15
  SPEAKER_00: I want basically a report with evidence of data that showed how I made my decision.
• 20:22
  SPEAKER_00: That's the governance level we're talking about.
• 20:24
  SPEAKER_00: And that's what Exonus Decision Intelligence does.
• 20:28
  SPEAKER_00: So it says, okay, here's all the data I have access to in my data space.
• 20:33
  SPEAKER_00: Here's data that's that I'm surfacing that, and we send we call them signals.
• 20:39
  SPEAKER_00: Here's signals, here's a data, a flood sensor that's nine feet over where it should be.
• 20:43
  SPEAKER_00: Um, that's now prioritized, right?
• 20:45
  SPEAKER_00: Okay, click on that signal.
• 20:49
  SPEAKER_00: Give me all the history of this flood sensor, the data around this flood sensor.
• 20:53
  SPEAKER_00: Give me, you know, history that shows other flood sensors nearby and correlates and makes sure that this is not a false reading.
• 21:01
  SPEAKER_00: Pack it, okay.
• 21:02
  SPEAKER_00: Here's all that data, right?
• 21:03
  SPEAKER_00: AI is pulling that data for you.
• 21:05
  SPEAKER_00: All right, let me investigate a little bit of that.
• 21:07
  SPEAKER_00: Now give me your sense of an analysis on what this data means.
• 21:12
  SPEAKER_00: The AI gives you that in a narrative.
• 21:14
  SPEAKER_00: Then you pin evidence to that narrative.
• 21:18
  SPEAKER_00: You do it, the human does it and says, I think this evidence, this evidence, this evidence should be pinned to this narrative.
• 21:23
  SPEAKER_00: I'm gonna tweak the narrative, and then I'm gonna um seal that and send that to somebody else to look at.
• 21:30
  SPEAKER_00: So I'm sealing it, I'm attesting that this is the evidence, this is the narrative that I believe in, and this is what we should do.
• 21:36
  SPEAKER_00: I'm gonna seal it cryptographically, and I'm gonna send it up an approval chain.
• 21:40
  SPEAKER_00: And then somebody else is gonna see it, open it.
• 21:43
  SPEAKER_00: Oh, look, some of the evidence changed.
• 21:45
  SPEAKER_00: Hmm, that's interesting.
• 21:46
  SPEAKER_00: Let's run, let's rerun the analysis to see if it gives us the same output.
• 21:50
  SPEAKER_00: Eventually you seal that, you make you call, you call the news organization, tell them people need to flee, and you have that now as a as a piece of evidence and data.
• 22:00
  SPEAKER_00: That's in the system that you can call back whenever you need to.
• 22:02
  SPEAKER_00: So that's what we do.
• 22:04
  SPEAKER_01: So it sounds like you need to have some kind of infrastructure around uh those governance roles, such as a data steward or an application owner.
• 22:13
  SPEAKER_01: Somebody needs to be able to say, okay, I'm looking at this data, I'm interpreting this.
• 22:17
  SPEAKER_01: So it sounds like there'd be multiple levels or multiple layers of uh consumption and decision making that feed into an overarching governance structure, right?
• 22:26
  SPEAKER_01: Because at the end of the day, you need to make decisions, you need to be able to back those decisions up, and you need to be able to, you know, collect that evidence and timestamp it and seal it and have kind of a chain of custody or or or a chain of value uh from that data, and then be able to uh justify what what decision you're making.
• 22:45
  SPEAKER_00: So um our role in that is to map to their actual governance, right?
• 22:49
  SPEAKER_00: We have configuration packs that basically you can very quickly change the pack and it updates the application based on your governance structure that you need.
• 22:58
  SPEAKER_00: And our role is to cryptographically seal the decisions and the evidence based under governance section.
• 23:03
  SPEAKER_00: Gotcha.
• 23:04
  SPEAKER_01: Uh something that we we do, and and just so you know a little bit about who early information science is, we've been around for a few decades, and we work in information management problems for global, you know, 1000 companies.
• 23:15
  SPEAKER_01: So uh we we've uh I I I uh wrote an art an article in 2016 called There's No AI Without IA.
• 23:22
  SPEAKER_01: I don't know if you've ever heard that phrase, but um, there's no artificial intelligence without information architecture.
• 23:28
  SPEAKER_01: Uh and that's uh kind of been an industry catchphrase for a while.
• 23:31
  SPEAKER_01: I wrote a book in 2020 called The AI Powered Enterprise, which really talked about ontologies and knowledge architecture as a way to uh support your AI initiatives.
• 23:41
  SPEAKER_01: Uh and even though, as before large language models, everything in that book is still valid because it really talks about the underlying structure.
• 23:48
  SPEAKER_01: When you go into an organization, are you working with uh any knowledge graphs or ontologies?
• 23:54
  SPEAKER_01: Or what's the role of uh information architecture and reference architecture in what you're doing?
• 24:00
  SPEAKER_00: So one thing you have to understand about us is our technical founders have 20, 30 years in DOD and intelligence community and kind of data intelligence and AI.
• 24:14
  SPEAKER_00: Um so uh they they know Palantir really well, for example, right?
• 24:20
  SPEAKER_00: And ontology is kind of what Palantir is making sure, right?
• 24:24
  SPEAKER_00: Um we are a schemaless architecture, so we can help discover schemas out of the data that we get into our system.
• 24:34
  SPEAKER_00: Um and so that means we can ingest anything, we can get to value pretty quickly for use cases, um, we're pretty resilient to change.
• 24:44
  SPEAKER_00: Um, you know, kind of AI native is the way we kind of think about that.
• 24:49
  SPEAKER_00: Uh but we can also inherit an ontology.
• 24:52
  SPEAKER_00: So if there is an information architecture, we can we can map to that.
• 24:56
  SPEAKER_00: So I think the way to think about what we do is we're flexible on that.
• 25:00
  SPEAKER_00: If you don't have one, we can help kind of shape one more dynamically.
• 25:05
  SPEAKER_00: Um, if you do have one, we can certainly map to that using our configuration path.
• 25:09
  SPEAKER_01: Okay, that's helpful.
• 25:10
  SPEAKER_01: Um, so uh there's something that we are you mostly working with government agencies, or do you also have corporate uh targets?
• 25:19
  SPEAKER_00: Yeah, the one the one I told you was actually sounds government, but it was a commercial um deal, and we have other commercial customers.
• 25:26
  SPEAKER_00: Yeah, and then um we have uh we are in the Department of Defense and Department of Energy from a government perspective.
• 25:34
  SPEAKER_00: I think the the the similarities are there's edge data, some sort of you know, data gravity that makes it hard for them to centralize or unreasonable for them to centralize, and that could be either there's three reasons actually.
• 25:47
  SPEAKER_00: There's technical reasons, like it's video, you just don't want to put it, you know, bandwidth constrained or something.
• 25:53
  SPEAKER_00: This the the next reason is regulatory, it's HIPAA or PII or something, and it's just resists centralization because you know you could do it maybe, but there's a lot of things you have to think through.
• 26:04
  SPEAKER_00: And the third one is political.
• 26:06
  SPEAKER_00: You can imagine an enterprise organization.
• 26:08
  SPEAKER_00: I was talking to the head of Gen AI at um a bank, and she was like, The hard part is we have these application owners in our IT organization who own their application and their data.
• 26:21
  SPEAKER_00: And they, you know, they're resistant to sharing it, and so you it's a political battle.
• 26:25
  SPEAKER_00: And I said, Well, that's cool because our architecture, you could give them a little node and they could own that.
• 26:31
  SPEAKER_00: It's a wholly decentralized architecture.
• 26:32
  SPEAKER_00: So they you could put one in front of something like that.
• 26:35
  SPEAKER_00: And so our technology maps to a political decentralized data problem as well.
• 26:41
  SPEAKER_01: Political reality on the ground.
• 26:43
  SPEAKER_00: Yeah, yeah.
• 26:43
  SPEAKER_00: Yeah.
• 26:44
  SPEAKER_00: So um, so yeah, I we're we're across, I would say we're we're heavy in government and we're heavy in sort of uh you know, edge AI type use cases, so sensors and IoT and things like that.
• 26:55
  SPEAKER_00: And but I think we're seeing enterprise use cases as well.
• 26:57
  SPEAKER_00: We have one that's just like, I got some data in this cloud, I got some data in that cloud, I don't really want to centralize it, and we're we're doing predictive stuff across that for them.
• 27:06
  SPEAKER_01: Well, we have a um, I I would be it'd be interested in getting your your feedback on something that we've just um launched.
• 27:12
  SPEAKER_01: It's uh an AI readiness um uh quick check.
• 27:16
  SPEAKER_01: Uh we have a big AI readiness, you know, 74 questions, 17 factor kind of consulting engagement.
• 27:22
  SPEAKER_01: But this is something that's meant to be a survey, somebody can answer in you know, 15 minutes, and it gives them a report that's based on our IP, our work in governance, our work in information architecture, or knowledge architecture.
• 27:36
  SPEAKER_01: And it grades across four different um domains knowledge uh architecture, knowledge engineering, knowledge readiness, operational readiness, uh, technical readiness, and then governance readiness.
• 27:47
  SPEAKER_01: And uh it's been something that people have we were giving it away for free, and someone said, I can't believe you're giving this away for free.
• 27:52
  SPEAKER_01: There's so much value here uh because it's based on all this IP.
• 27:56
  SPEAKER_01: I would love to uh uh send you the survey and uh have you, if you're interested, have somebody in your organization take it, send you the report, and just get your opinion on it.
• 28:05
  SPEAKER_01: Because again, it's something that we're kind of beta testing and pushing out to be who knows, maybe be of interest to some of your targets or prospects, because what it does is it uh it identifies some of the gaps that they may not be aware of, you know.
• 28:18
  SPEAKER_01: Yeah, and it might end up being, you know, a diagnostic tool you can use, it might be an awareness tool, an education tool.
• 28:24
  SPEAKER_01: So so if you're open to it, I can send you that um survey and then we can write the producer report.
• 28:29
  SPEAKER_01: We've been charging$2,500 for the first report, a thousand for each additional, but I'll just comp it to you if it's something you'd like to take a look at.
• 28:37
  SPEAKER_00: Okay, sure.
• 28:38
  SPEAKER_00: We'll take a look.
• 28:39
  SPEAKER_00: Yeah, yeah.
• 28:39
  SPEAKER_00: I'm not sure we, you know, we're a small little company, right?
• 28:42
  SPEAKER_00: So we can give you feedback on the survey.
• 28:44
  SPEAKER_00: I'm not sure what a great results.
• 28:46
  SPEAKER_01: Well, why don't you take why don't you take it and see what the results are?
• 28:49
  SPEAKER_01: Yeah, that'll be fine.
• 28:50
  SPEAKER_01: It doesn't matter what how big you are, it it could be done at a department level.
• 28:53
  SPEAKER_01: So yeah, you know, and then uh then that'll give you might even give you some thoughts on where you need to go internally, right?
• 29:00
  SPEAKER_01: Calling children is always a challenge.
• 29:02
  SPEAKER_01: All right, I know we're at the right at the bottom of the hour, but um what we'll do is we'll write up uh kind of a structure for this, some questions.
• 29:09
  SPEAKER_01: Uh we'll uh run that by you, and you can add any details to that that you'd like.
• 29:14
  SPEAKER_01: But again, like to start with a problem statement or or misconception, and then what to do about that.
• 29:20
  SPEAKER_01: And that's a good way to kind of hook people and then and then go from there.
• 29:24
  SPEAKER_00: Uh sound like a fit for what you guys are doing?
• 29:27
  SPEAKER_01: I think so.
• 29:27
  SPEAKER_01: I mean, I I think it's um I think that I'd like to understand a little more how you differentiate in the marketplace um uh in terms of because again, a lot of organizations are trying to do that or you know, talking about this data problem from that perspective.
• 29:44
  SPEAKER_00: Yeah.
• 29:44
  SPEAKER_01: So maybe a little bit more about um, you know, what's really different?
• 29:49
  SPEAKER_01: Like what are what are some of the other ways, what are the conventional ways of looking at this in the industry?
• 29:54
  SPEAKER_01: And I think you talked about that.
• 29:55
  SPEAKER_01: I think you talked about the fact that you have the ability to take that snapshot in time, to collect those signals, to package them so that they become kind of a decision-making object, and then that gets passed along to uh decision makers and and and gets surfaced in the organization.
• 30:13
  SPEAKER_01: And you know, if you want to compare that to kind of other ways of doing it or other approaches that you've seen, and talk a little bit about what the drawbacks are.
• 30:22
  SPEAKER_01: So we want to do is we want to give people kind of something, something to hang their hat on in terms of an a new way of thinking about this and innovation, what's different, right?
• 30:31
  SPEAKER_01: It's kind of like the so what perspective.
• 30:33
  SPEAKER_01: But I think I think it's a a reasonable conversation to have, and I think it's uh you know, a certainly a problem that people are are struggling with.
• 30:41
  SPEAKER_01: Yeah, does that make sense?
• 30:43
  SPEAKER_01: Yeah, all right.
• 30:44
  SPEAKER_01: All right, great.
• 30:45
  SPEAKER_01: Thanks, Sad.
• 30:45
  SPEAKER_01: Nice to meet you.
• 30:46
  SPEAKER_01: Yeah, nice to meet you.
• 30:47
  SPEAKER_01: I'll shoot you that survey information.
• 30:49
  SPEAKER_01: You can take it and uh and then you know we'll send you the report, and I think you'll be impressed, and at least it'll it'll give you something interesting to to think about.
• 30:56
  SPEAKER_01: Thank you.
• 30:57
  SPEAKER_01: Yeah, nice to meet you.
• 30:58
  SPEAKER_00: Looking forward to it.
• 30:58
  SPEAKER_00: Nice to meet you.
• 30:59
  SPEAKER_01: Okay, take care.
• 31:00
  SPEAKER_01: Yeah, bye.