Benjamin Tosado: Welcome to Cleared for Takeoff, real AI stories from experts in industries that move the world. Hello, everybody. Welcome back to Cleared for Takeoff. I’m Ben Tosado, CEO and founder of Deep Blue Cloud Computing, your host. And today, we have Jon Baker from Airvoyant, a company that’s doing something completely different in aviation procurement.

They’re building AI agents to solve some of the biggest and longest standing challenges in in aviation parts procurement. They’re also building an AI foundry, which is as interesting as it sounds. So I’m very happy to have you on today to talk about this stuff, Jon.

Jon Baker: Thanks, Ben. Good to be here. Good to be chatting with you.

Benjamin Tosado: Yeah. Yeah. Absolutely. So before we jump into all of the really cool things that are that are going on at Airvoyant, can you tell us a little bit about your background in the industry and what got you interested in integrating AI into your into your platform?

Jon Baker: Yeah. So my my background is I was a software person for a long time, software engineer, architect, Started that journey at a long age. Went from building products to doing consulting about halfway through. Did work in a lot of experience design and marketing and bringing digital to a lot of Fortune 500 brands. Along that journey, Boeing actually became a client of mine.

That’s actually where it really And, we did a lot of work for them. That was fantastic. Super interesting. Then we did some work for the aftermarket, and that’s where I first started coming across parts. And then there was a little bit of a break in the action.

I ended up with another job working for Big Five Consulting and got into aftermarket hit again. It’s like it was meant to be. I was I was drawn. It was like a magnet. I couldn’t get away from this thing.

Right. And so ended up getting into parts, procurement again and doing some consulting work, mainly for the OEMs at that point. So their this and efficiency. And so I learned a lot from that. That was that was a very important chapter.

And then, you know, coming full circle to today ended up connecting back with friends at Trax and AR, and they were interested in creating a modern software platform, taking advantage of everything they can to transform parts procurement. So that was kind of the professional, you know, the professional journey that got me from the beginning to where we’re at now. And then personally, aviation has just been, it’s been part of my life since I can remember.

Benjamin Tosado: You’re actually a pilot, right?

Jon Baker: Yeah. Pilot. And I got various family members that were flying. My dad flew sail planes. We flew model airplanes together.

So it all started when I was very small and hasn’t stopped. So it took a while, but eventually professional interests and my personal interests did come together there. And here we are today with Airvoyant.

Benjamin Tosado: Yeah. Yeah. No. That that’s awesome. It’s it’s cool.

Actually, as much as we’ve talked and worked together, I didn’t know that you you had, you know, the the work with Boeing, and and I knew that you did a little bit of of of piloting and stuff, but I didn’t know aviation was such a big part of your of your history. So that’s really cool.

Jon Baker: Yeah. Yeah. Big big big part of it. And so far, that’s worked well. Sometimes when people ring those two things together, it’s not always a great combination.

Right? One ends up being disappointed in some way, but that that was that was never the case. I mean, certainly, when it started to work at Boeing, I was living in Seattle, was starting a new division up there for the company I was working for. And, you know, suddenly, I’m down there in Renton where seven three sevens are being made, standing in the manufacturing base, and it’s like, you couldn’t believe it. Was just amazing.

And then seeing aircraft taking their first flights, it was awesome.

Benjamin Tosado: That is awesome. And cool that you’ve got the software background and now working on these AI projects in an industry that I really think needs the help. Right?

Jon Baker: Mhmm. Definitely.

Benjamin Tosado: So so can you talk a little bit about the problem that that you’re solving with Airvoyant and and how it’s gonna how it’s gonna help an industry that’s under kind of increasing pressure from a variety of different things?

Jon Baker: Yeah. I mean, parts, if you look at our airline customers, just to center everyone as to who we serve, parts and procuring parts is one of the big expenditure areas, certainly not quite as much as fuel and a few other things at the very top of the list. But boy, it’s it’s right up there. And that makes sense. Aircraft are expensive.

Aircraft parts are expensive. You’ve got to find them. You’ve to get them to the right place on time.

Benjamin Tosado: Yep.

Jon Baker: And what happened is that when we started looking at this and I got an early early signals when I was involved in parts a little bit more on the seller side is how manual the processes are that people go through to managing that procurement process. Systems that are not connected. A myriad of techniques that they might apply to go and find material because it’s just hard to locate what you need when you need it. The vast network of vendors, thousands of them all around the world who are all basically at different levels of sophistication and integration. And when you stir all of that up, what you end up with is an incredibly inefficient ecosystem that has basically been that way for a long time.

And, you know, we saw various people taking a go at trying to improve it right over the last even twenty years. But it’s never quite, you know, broken free and created that scalability and connectivity that I think people truly need. And we just felt that, you know, the way technology is moving right now, this was an opportunity to change that. This was this was the time where we could get beyond this limitation that we’ve seen in the industry and actually start to take it to that next level. And we’ll see whether we we are going to be proven right here over the coming months and years.

Benjamin Tosado: Yeah. Yeah. Let me ask you this about something you said. Why do you think it hasn’t been solved? It’s been a problem for so long, right?

Jon Baker: And

Benjamin Tosado: it is a big expense area for these aircraft operators. Why do you think nobody’s been able to solve it?

Jon Baker: I think it’s been a little bit underserved. So when you think about other areas, let’s stay in our MRO since that’s a bit more of my lane here. There’s other, obviously, other aspects of airline operations that got some similar challenges. But I think it’s been underserved because it’s not always perceived as that flagship problem. Right?

We see a lot of people have tried to solve. Let’s provide intelligence systems to try and help solve some of the maintenance challenges, bring this knowledge base to life Right. And put it in the hands of mechanics, people turning wrenches. And I think those use cases were often higher, higher profile. And so procurement was left a little bit behind.

It was seen as, oh, yeah, you guys buy the parts. Well, in general, they kind of show up. So, you know, we’ll just keep moving on. So I think it was a little bit of under investment in general across the board. But I think that this fragmentation of the vendor ecosystem also is a challenge.

It’s a challenge that no matter what you’re gonna do and what technology you can apply because you’ve got everything from the larger OEMs down to small mom and pop vendors who might still have certain materials that even large airlines need just for specific aircraft. So that makes it that makes it very difficult to bring all of those folks together. You know, we haven’t seen an Amazon like effect on our industry. You know, Amazon in many other areas, whether it be consumer and biz and various industries and businesses provide this kind of normalized central hub for conducting commerce. We never saw that here in, parts procurement in aviation.

And, yeah, certainly aviation’s got its own different challenges. Right? The just the width and breadth of the material is highly varying. We got things that, you know, are huge crates with components that cost tens or hundreds of thousands of dollars all the way down to nuts and bolts and oil and things like that. So you you’ve got this wide gamut of material that you’ve got to handle and manipulate.

Right. You’ve got a very fragmented vendor system. And then you’ve got customers who’ve also got the allied side, who’ve also got a highly variable set of platforms and technologies and a very limited amount of bandwidth to go and necessarily make, you know, bigger steps, progressive steps in improving things. So you put that all together, and I think you you you see people chipping away at the edges. Right.

And you see some some improvements. I I definitely do not want to dismiss some of the great things that have been done, but they just haven’t created, you know, a seismic shift in what’s going on. It’s just making pieces of it a little bit easier, but we haven’t solved the whole problem.

Benjamin Tosado: Right. And then you’ve got the whole compliance aspect of it too. Right. That adds another layer of complexity.

Jon Baker: Yeah, sure. I mean, just certainly all of the paperwork, no one in the procurement game or listed in the parts game will say that just buying material is the only use case. There’s a whole set of use cases around exchanges, loans, borrows. And then, of course, the big the other big elephant of repair.

Benjamin Tosado: Right.

Jon Baker: You know, they’re all they’re all coming together there. They’re all involved in that day to day activity. And again, they’re all suffering some of those same complexity. Certainly when you get into the repair business, you’ve got, you know, a lot of documentation and content that you have to go through and understand. And that gets really difficult.

Right? And that is one instance of where I think the current generation of technology that we’re now digesting and applying to our solutions probably will be applicable here going forward for repair.

Benjamin Tosado: Yeah. Yeah, absolutely. So that’s a great overview of the problem. So let’s talk about Airvoyant, right? So this is super exciting.

You guys went in and launched the platform at MRO Americas some weeks ago. We were there with you to do that. Very excited about that and working with you on the project. So tell tell us about Airvoyant, where you started, how you decided to pivot, what you’ve what what you’ve built so far in terms of AI agents in the system, and then what’s what what what kind of the road map looks like?

Jon Baker: Well, Airvoyant is our vision for automating parts procurement. And, you know, it’s a it’s a pretty audacious goal, but we believe that there is line of sight to providing a fully automated ecosystem. We focused on the buyer side. I think a similar effort efforts will happen on the seller side.

Benjamin Tosado: Yeah, I agree.

Jon Baker: And we, we believe that the reason that automation can now succeed is obviously because of the incredible progress that we’ve seen with AI and all of the capabilities that are now emerging. So that’s if we go to the current story right now, that’s where we’re at. In terms of the journey, yeah, that’s not where we started. We started a little bit more traditional. We thought that, I was referring to Amazon a little bit earlier.

We thought that an Amazon like Marketplace probably was the solution because obviously we’d seen an Amazon like effect in all of these other areas of selling things, whether it be consumer based, business based, very successful. And so we thought, oh, well, maybe that’s what we need to go build. We’ll just bring everyone together in this marketplace, and, of course, that would be producing a solution again for humans to sit there and run their searches. We we could, you know, integrate with various back ends and, you know, bring in material, lists that they’re looking for and not have people typing in bar numbers and things like that. So if you assume that that would be doable and we can have vendors loading inventory just like, people do in in Amazon today.

So create that type of marketplace, and that would be the solution. And, of course, while we’re doing this, this is a couple of years ago now, while we’re while we’re brainstorming this and trying to basically figure out what this would look like. Yeah, the world’s kind of shifting underneath. And, of course, when did ChatGPT come out? I think it was the end of twenty twenty two.

So we’re we’re coming up on the fourth.

Benjamin Tosado: That’s when the genie came out of the bottle.

Jon Baker: Right. Came out of bottle. It’s definitely not going back in the bottle. Yeah. That’s for sure.

Yeah. The genie came out of the bottle in 2022 and obviously a huge huge interest to us. I mean, that we we weren’t ignoring it or anything. But what I couldn’t quite see was at the beginning was how that was going to impact our approach. And so two things happened.

First, the first thing was, of course, I’m working with my friends here at Trax who own a large MRO ERP, one of the big three that airlines typically use. And for those who aren’t familiar with that type of solution, those those ERPs basically for an airline, it just stores all of your information about your your aircraft, your people working on the aircraft, your parts, your compliance, even all the way down to your ground service equipment, like everything in your MRO operation is in there. And that’s what you use to keep everything straight. That’s your system of record. And we we looked at connecting to that.

We thought, Well, you know, if you’re gonna buy material, you know, that’s where it starts. Right? So if you’re gonna perform a maintenance activity, it could be scheduled, it could be unscheduled, but doesn’t matter. Whatever it is, whatever you’re gonna do, it starts there and your inventory is there. So okay.

That’s great. So we know what we’ve got on hand. We know what we’re meant to be doing. We can end up basically bringing all that together and you end up with, okay, what’s this the stuff I need to go buy to go, you know, do something? So we started off going down that path, and we thought, wow, that’s actually really interesting.

So if I can just actually take that list and actually go and start to search the market, what what do I need to do? I just need to connect the vendors. And there are vendor networks out there today. There are ways of, you know, looking at what vendors have, and getting prices and requesting quotes and things like that. So we thought, wow, we’ll actually just start connecting these things up.

We don’t really need to create our own central marketplace. We’ll just pivot here a little bit and skip a few things and just start to to to bring, you know, buyer and seller side together. And we’ll do that with a direct integration with the ERP. Obviously, starting with tracks, we might support some others later. That’s fine.

And so that that was very promising. We started building that. But what while we were doing that, of course, AI is moving rapidly. And, you know, this the change of events was when the AI crowd went from being super excited around LLMs and generative AI, which obviously was fantastic and amazing start, and then how they transformed using that type of technology to produce agents. And agents are a way of producing software that is running, that is able to take on tasks and more importantly, actually make intelligent decisions.

Right. So we’re working on what we think is the right solution and is the direction we’re gonna we’re just gonna, take the data straight out the ERP. We’re gonna hook to the vendor networks. We’re gonna conduct some transactions and parts will be flowing. Yep.

We had to do that part. But what we were missing was, oh, yes. And we could now potentially build an agent workforce that could actually run on top of that type of capability and start to make decisions. And we thought, okay, is this a good use case for that? Are we gonna what types of decisions are we gonna make?

Is the technology gonna support that? And it became very obvious to us when we started looking at this, and I did a lot of work with the folks at Amazon, which by the way, big shout out to that, my my team there, because they were advisers in the early days who gave us, I think, a little bit of inspiration because they showed us, you know, that art of the possible and what they were doing. And, they looked at our case and said, well, Jon, I think that this this is probably a pretty good use case. And the reason why is if you look at a lot of the decisions that have to get made in procurement on a day to day basis buying aircraft parts, and there’s a lot of them, But there there’s something that you can model. They’re they’re not they’re not mission impossible to go and get your arms around those things.

You can model them, and therefore you can potentially put them into something like an agentic solution. And now, you know, that kind of goal of trying to get to a higher level of automation was suddenly, we’re suddenly right there. But, you know, the challenge was we were halfway through the project. Right. Now, would argue we didn’t do anything.

We hadn’t built anything that was necessary contrary to putting an agentic workforce together to do past procurement. In fact, think most of the stuff we built was absolutely required. It’s the foundation of of providing something for the agents to run on top of. But, boy, it was a pretty sharp pivot. We had to change the change the project, move things around very, very quickly.

And then, of course, you’re working with, you know, your fantastic team that brought in all of that expertise on how you would go about doing this. And if that all wasn’t enough, at the same time, we are now digesting AI assisted development of the solution. Yeah. So yeah, I’d like to partition these things for people who don’t live in this world. I’m trying to build an AI solution.

Right? Bunch of agents who are gonna do procurement and potentially some other supply chain things. Great. The best way to get there is to use AI to go create that.

Benjamin Tosado: Right. Right. We could we could be doing all that manually. Right?

Jon Baker: Could be. Sure. Could be. Right. You could sit there trying to trying to write the code manually.

That’s insanity, by the way, definitely don’t want to do that. And, you know, we, we know that this is the right way to do it because if you look at the AI software platforms, whether it be an OpenAI and Anthropic or any of the others, they’re all using AI to develop AI. And so we had to go and bite that off. It was around the time we were making this decision on Agents, a ClaudeCode had just come out.

Benjamin Tosado: Yeah. Exactly. All of these great tools. Right?

Jon Baker: Yeah. Yeah. And so I thought, okay. I’m gonna I’m gonna put my developer hat back on. I’m gonna do a project.

And I actually my first project was to try and build a quick simulation of our service in Airvoyant. Just a quickie. Right? I’m gonna build a little web application. I’m gonna simulate parts coming in, parts requests coming in.

I’m gonna simulate a little vendor network. I thought, oh, yeah. Yeah. We’ll never be able to do this. This is gonna be too it’ll prove to be too much.

I can tell you about an hour and a half later, I had a pretty good starting point. I knew nothing about Claude code. I put the plug in into Visual Studio, provided it up, started asking questions, and it started creating stuff.

Benjamin Tosado: Yeah.

Jon Baker: And I can remember I ran out of tokens pretty quickly. So then I had to bump up my subscription, which I did. Right. I carried I carried on. And I remember I remember going in into my wife.

I think she was sitting in the living room or something saying, my god, you believe what I’ve just created this afternoon. This thing’s amazing. I love new technology. I’ve been been all over new technology in my career since it started when I was a teenager. So but I’m also pretty skeptical of things, and I know the pace at which it typically happens.

I’d never seen anything like this. I mean, this is just like it was like magic, but this Transformative. Yeah. That’s a big that’s a big there’s a few big days in your professional career that you remember. I’ll remember that one.

I know I’m probably near the end of my career, than the beginning, but, boy, that was just a big day. And so now we know that we’ve got this agentic technology coming. Now we know we’ve got AI assisted. I had a colleague of mine. He was he was running similar experiments, and we were kind of sharing war stories.

And I got other friends who’ve now retired out of software development. They were doing similar things. We’re all just correlating the same thing. Like, god, can you actually believe this thing?

Benjamin Tosado: Right.

Jon Baker: This is absolutely incredible. So now I knew that we had to do this because we could have waited. We could have said, you know what? Okay. Let’s just tap the brakes.

Let’s finish off what I called workflow automation, which was kind of the base platform. Let’s just finish that off, get that out, test it, and then we’ll come back and do the agents. And it was very obvious that was not the right approach. We we didn’t have time because when you join the dots together here, if we can start to build agents with AI, and we can do it and we can learn to do it, which takes a while, it’s not like it happens instantly, but you start to develop it. If I can do that, it means other people are gonna be able do as well.

So we just can’t we can’t wait six months and sit down in a conference room and review and pontificate over the State of the Union. And how well did, you know, our workflow automation go? Nope. Absolutely not. We had to go and start putting that together as soon as possible.

And, you know, here we are. We we we launched with, our first agent. We’ve got line of sight to three others fairly quickly. In fact, we were working on them this week intensely, and I think more to come. And the speed at which we have to develop those is is just it’s it’s got to be fast.

We just can’t we just can’t stop. You just gotta keep going here. And now we’re gonna get into the real learnings of building agents and what that means and how you build them up over time. You you there’s some good lessons learned already that you wanna start fairly small, get small things working well, then combine those small things together and start to build it up from there. Sounds logical now, but sometimes when you get into this stuff, you go, wow.

You come up with a big grandiose ideas and you wanna shoot for

Benjamin Tosado: the fence right at the right at the beginning, but that’s not quite the right way to go. Anyway Sure.

Jon Baker: That that was kind of the trilogy.

Benjamin Tosado: Got it. Yeah. So we we’ve completed the sourcing agent for you guys. Mhmm. Three other ones I know we we’ve been doing design sessions all week.

Can you talk a little bit about the sourcing agent? I think ordering agent’s gonna come next. The other two that are immediate on the road map.

Jon Baker: Mhmm.

Benjamin Tosado: And and just give our listeners a little bit of an idea of what these what these specific agent workers in the workforce do.

Jon Baker: That’s the agent that’s actually gonna take the material requests, and it’s gonna search the vendor network. It’s gonna get responses back. Who’s got material? What are the prices? There are catalog prices.

Great. We’ll receive those. If we issue RFQs, we’ll get quotes back. And then it’s gonna perform the analysis. So let’s say I get a handful of responses for a particular piece of material.

Maybe it’s an expensive piece. Maybe it’s difficult to find. So we’ve got probably a fairly complex task here to look at those responses, not just, you know, price and lead times, but even looking at do we receive documentation? Do we have the eighty one thirty if you’re over here in, The US? Things like that.

And that she makes some recommendations. And the agent is going to not just use the data in the responses that we received. It’s also gonna look back at the historic data. So one of the things that Airvoyant does is it has a big data lake. And so every event, every transaction, anything that happens in Airvoyant between the buyer and the seller we put into the DateLake.

That’s our historic record of everything. And so with obviously, it builds up over time and gets gets larger and larger. But at any point in time, the sourcing agent can go to that data lake and we can actually go and perform analysis, perform some data science. That’s not AI or agentic per se. That’s just good old fashioned data science and man.

But we can do scoring, right? We can go and look at our history with that part. How many times have we ordered it before? Where did we get it from? What we pay for it?

Did we buy new? Did we buy used? What’s our history with PMA? And we can also tune that type of analysis a little bit for the specific customer. Everyone’s got their priorities and where do they want to emphasize the importance in making that decision.

But the agent basically is going to make recommendations. And with the recommendation, it also is gonna give us a confidence score. So we’re not asking the agent to just come up with a recommendation and that’s it. Just leave it at that. There are going to be some situations where it’s very confident because we have a lot of data, lot of history, or let’s say in the responses, it’s just fairly obvious which one is the best one.

There are going be other situations where that’s not true. Maybe we’re looking for something a little bit more esoteric. Maybe we don’t have a lot of data around it. And so the confidence score is probably gonna be less. And we want the agent to be honest about that and just say, you know, call it as it is.

But what will happen here over time is buyers will make their purchase decisions. We’re still allowing human in the loop at that point.

Benjamin Tosado: Mhmm.

Jon Baker: And human in the loop is going to look at those recommendations. And if it chooses our recommendation, that’s great. We’re off to the races, place the order, move on. If it does, if they don’t choose our recommendation, then we ask for feedback. Why didn’t you choose it?

Simple set of checkboxes, check a couple of things, but that goes back into the training loop of the solution, and we can get a little bit smarter. And over time, we will be able to record those decisions. The history in the data lake will build up. And overall, while there will be some situations where confidence levels will just not be high, because again, you just don’t have a lot of data or it’s a very esoteric type of transaction. I think for the most part, majority of the transactions will have high confidence levels.

And the reason that we’re pushing so hard on this thing, and I’m I’m I’m kind of emphasizing it, is because when you want to get to the point where you say, okay, this thing seems to be working reasonably well. I’m I’m always choosing your recommendation when the confidence level is above a certain percentage. Why am I pushing the button? Why don’t you just order it? Right.

So it’s pretty obvious. Let me deal with the ones that are below that level. Once above that level, you take and place the order. So that’s a good example, I think, of building an agent. And we’re working our way up to that next level of automation.

And the thing about it. Yeah, you do need some historic data. We do need to, you know, run the system for a period of time, but it’s not like you’re going to run it for years. When we start with customers, we’re going to ingest historic data. It’s not like I have to sit here and wait for a year to get a year’s worth of data.

I can go. It’s it’s usually in the ERP and other systems. I can pull that data in there and we can get a good starting point. But that’s, I think, how the sourcing agent’s going to start. And then we start to work on the rest of the workflow.

So you mentioned ordering, and I just mentioned placing orders automatically. The sourcing agent will talk to an ordering agent, and the ordering agent can make can start to make decisions. Is this something I should order automatically? Is this something I should push back to the humans, human in the loop? And you can imagine there’s just a whole plethora of criteria that you could potentially put in there so that everyone has the right comfort level.

We’re not pushing automation on everyone. We’re providing automated capabilities. Our customers will choose how they want to use it, when they want to use it, and some of the business rules and logic that will go behind that. But we will have an agent that will have that will be configured for that and can make those decisions. And then you can go either end of this life cycle.

We can go upstream. We can look at the list of demand coming into the solution. Do we actually need to purchase all this stuff? I’ve heard that from a number of customers who claim they’re buying too much. They’re they’re buying too much.

Benjamin Tosado: Too much material. Right. Yeah.

Jon Baker: And my solution is probably gonna accelerate that because now it’s very easy to buy stuff. So we don’t wanna do that. So some upstream control and optimization, maybe a little bit of forecasting. We’ve got some potential agentic functionality around, the customers’ the customers’ inventory. What have they currently got, or or what are they gonna get?

Because, you know, when you’re doing long term planning exercises and you’re starting to get those material lists for heavy checks, that that could be months in advance. We don’t actually need those parts today. We’ve got a little bit of time where we can start to think about when is the best time to go and potentially purchase that material and what impact are we gonna have on inventory during that time? This is why I think a lot of people end up with too much material is that this is a bit of a time based problem, and things are very dynamic and changing. So today, you might look at it and go, oh, yeah, we’re gonna do that heavy check.

We’re definitely gonna need to buy these things. And in the past, this has been a pain to go find these parts. Wondering if you guys go buy it. Whereas in fact, if it waited a little bit, we may find, hey, we’ve got some exchanges. We’ve got some repairs coming back.

We’ve got other situations here that are changing our own inventory. Maybe we don’t actually have to buy this material, but we didn’t wait. We just took an easy path, but it generally resulted in too much stuff. So, you know, these are examples of other agents and I think there will be some big ones like that, demand agent. I think there’ll be some smaller ones, ones that might be looking at contracts, LTAs, and looking at the value that those are providing.

Is this a good decision if we’re coming back if we’re coming up to a milestone where we have an opportunity to renegotiate that contract, what should I be asking for? Am I are they is that vendor delivering the parts on time? Are we getting are we achieving the terms that we set out into the LTA? Is it delivering the value that we want? So, again, we can go on and on here.

There’s probably a dozen or so different types of agents, I think, within what broadens from just starting off with procurement and gets into a wider supply chain.

Benjamin Tosado: Yeah. Absolutely. Across the whole kind of procurement domain, like, that that’s that’s fantastic overview, Jon. Deep Blue has been working with you guys since the core of Airvoyant, one platform supporting you guys with some different things there. Helped develop these first couple of agents.

Now, we’re working on an engagement to develop more agents for you guys. And then the Foundry. Right?

Jon Baker: Mhmm.

Benjamin Tosado: So can you talk a little bit about the partnership with Deep Blue? How it started? How it’s evolved? And then talk about the Foundry. I’m really excited about it.

We’ve got a variety of other customers asking about the same type of capability. Right? Right. For somebody that’s never heard about AI Foundry, if we could talk a little bit about that too, that’d be awesome.

Jon Baker: Sure. Well, of course, we were joined together, I think, through our friends at AWS. Shout out to Amazon for for linking us up. That was, that was the perfect partnership. Yeah.

I mean, your team have been a lot of the, brains around how we have developed our AI agentic solution, which has been fantastic, great team, and also that AI assisted development critical to what we’re doing going forward. We’ve got to you’ve got to be on that train. There’s absolutely no alternative. And so if we look at how that is going to evolve here going forward, the interesting challenge is knowing how to scale a team long term to support these types of agentic capabilities. Now the general because it moves so fast.

And by the way, the technology is also improving underneath our feet. So it’s not like it stopped and we’re in a static period here.

Benjamin Tosado: It’s going faster and faster. Right?

Jon Baker: Quite quite the opposite. Yeah. It’s actually probably going to get even more dramatic is you have to structure a team that’s not too big and deep that is fully taking advantage of AI assisted development, but has enough throughput that we can put new agents out there and release them because we have to go back to, you know, a little bit of that principle I was talking about earlier where you start small. Don’t don’t take on too much too too quickly with these agents. Let them build their sophistication up over a little bit of time and a little bit of experience of running them.

And so building this dynamic team that can start to stamp these out, put them out there in the world, maintain them, and develop others as we go along. And this is kind of this foundry concept. When it’s not unique to us, we didn’t invent this. We see this in we see other AI companies that have done this. So we’ve kind of taken that same principle.

But the teams are fairly lightweight. They’re high powered. So we have very experienced people, but there’s not dozens of them. You know, I came from teams when I was working in software products and hardware where, you know, we had program teams of 200 developers. No problem.

Benjamin Tosado: Yeah. Yeah, absolutely.

Jon Baker: And, you know, whether it be HP, whether I was working with teams at Microsoft and other places like that, absolutely routine. I just, I just don’t think that’s the makeup anymore. I think it’s much smaller teams with heavy AI assisted development and figuring out how you are gonna use agents to develop your software and how you’re gonna harness them and control them, how you’re gonna feed them the right requirements, and as the solution builds, how you’re gonna vet those solutions, QA them, test them, deploy them, release them, which, of course, AI can help you with all of those things as well. So this is kind of the the world that we’re in right now. Your team is instrumental in helping us put this together, but that will be our our long term approach because we do have this, you know, this long list of agents that we’ve got to build and put out.

And I don’t see it stopping. You know, even with great AI assistance, humans still have to be there. We still have to figure out what it is we want to do. We have to go and get the tools to go build it. We have to go and make sure that what they built is what we want.

We have to join them together, operate them, evolve them. But it’s a different software life cycle than before. Not that we won’t have people, you know, touching code and writing lines of code and doing doing things. It’s not like we’re all sitting here with our arms crossed just staring at a screen and seeing what’s getting produced automatically. There’s still a tremendous amount of work to harness this, but the output is amazing.

You know, I think a team of three or four people today in an AI foundry is probably going to have the equivalent output in years by of, you know, 10 people.

Benjamin Tosado: Yeah.

Jon Baker: And I’m just based on where we’re at today. If we look at where we’ll be at six months from now, twelve months from now, you know, where’s that where’s that gonna end up? Because a lot of the keep trying to describe this to everyone. This technology is fantastic. It’s amazing.

But it’s very raw. Right? We’ve got very raw tools that we’re all figuring out how to assemble, and people are putting out great, great scripts and YouTube videos, and we’re all trying to get the best out of it. But eventually, you know, these platforms will be more refined. They’ll be easier to interact with.

They’ll be easier to get solutions out there, correct the first time. And so, again, I think this thing accelerates dramatically going forward.

Benjamin Tosado: Yeah. No, I couldn’t agree with you more. We’re seeing it across the board. I agree with what you said earlier too. We talked about a deep blue hour, our modern software development practice.

I don’t think you can have a modern software development practice without this now.

Jon Baker: Mhmm.

Benjamin Tosado: Right? Like, this is a fundamental part.

Jon Baker: I just remember that day I turned out to a cove for the first time. I haven’t I haven’t come down since. It’s like, oh my gosh. This is just amazing. So, yeah, I can’t think of any software team that and, yeah, everyone I know in software is looking at all of this.

Benjamin Tosado: Oh, yeah. But it is a bit of

Jon Baker: a question here. If you’ve got an existing big piece of software and you’re trying to bring AI assisted development into it, know, there’s an adaptation that you got to go through here. We were lucky. Right? We were greenfield, and we happened to be developing an AI solution and AI assisted development to build AI solutions absolutely made for it.

So we were we were in a very good position here to get that started. It’s it’s super fun. I gotta say if you’re a software person right now, it’s a very, very fun place to be.

Benjamin Tosado: Yeah. Throughout my career in the industry, I mean, feel like we’re at an inflection point right now. Right? I think the last the last big one was probably the transition to cloud technology. Right?

Mhmm. Like, 15 ago. Sure. I think we’re we’re at the beginning of a new cycle right now. That’s gonna be really, really exciting.

Right? So it’s really cool part to be a part of it and a really cool thing to be partnered with you guys on it. We really appreciate that.

Jon Baker: It’s a fun ride for sure.

Benjamin Tosado: Yeah, absolutely. So we’re almost out of time. Jon, why don’t you tell our listeners where they can find out more about Airvoyant?

Jon Baker: Airvoyant.com. It’s all there. I’m on LinkedIn. People will find me Jon Baker. Look at Jon Baker, J-O-N, and Airvoyant.

You’ll find me there as well.

Benjamin Tosado: Okay. That sounds awesome, Jon. Well, thank you so much for your time. Thank you for joining us today and talking about all the cool stuff that we’re we’re working on together.

Jon Baker: Thanks, Ben.

Benjamin Tosado: Alright. Bye bye. Thanks for listening to Cleared for Takeoff. Hit follow whenever you’re listening so you never miss an episode with change makers in aerospace, defense, and aviation. And to continue the conversation, find me on LinkedIn, Ben Tosado.