Intelligent Product Data — with Leo AI and OpenBOM

Episode Summary

The episode "Intelligent Product Data — with Leo AI and OpenBOM" delves into the integration of artificial intelligence (AI) across the product lifecycle management (PLM) domain, featuring Oleg Shilovitsky from OpenBOM and Maor Farid from Leo AI as guests. Both companies are pioneering in leveraging AI to enhance engineering productivity and design processes. Oleg Shilovitsky shares his journey from a mechanical engineer to an AI researcher and founder of OpenBOM, which focuses on optimizing product data management through intelligent algorithms. Maor Farid discusses how Leo AI uses advanced AI techniques to assist engineers in making smarter decisions during the design phase, aiming to reduce risk and accelerate time-to-market.

The episode highlights two critical insights: first, the role of AI as a tool for engineers to become more productive rather than replace them; second, the importance of providing relevant data to enable informed decision-making. Oleg emphasizes that AI should augment human capabilities by automating repetitive tasks and offering intelligent suggestions based on specific project requirements. Maor advises engineers to embrace AI tools like Leo AI to stay ahead in an evolving technological landscape.

For PLM and engineering professionals, the key takeaway is to integrate AI solutions thoughtfully into their workflows. By doing so, they can enhance productivity, reduce errors, and innovate more efficiently. The episode underscores that while AI presents new opportunities, it also requires a strategic approach to ensure its benefits are fully realized without compromising on human expertise.

---

Full Transcript

Michael Finocchiaro

Okay, we're live ⁓ and streaming. So this is Michael Finocchiaro I'm very pleased to introduce Oleg Shilovsky, who's one of my favorite guys on the web because we're always on podcasts together. And of course, Maor Farid of Leo AI, a very exciting startup. And we're going to talk today about AI across the product lifecycle, in particular how Oleg and Maor have used AI to produce really breakthrough, amazing applications. ⁓ So I'll let ⁓ Mauro, you can introduce yourself in the know like.

Maor Farid

Sure. First of all, ⁓ Michael, thank you so much for having us today and for the opportunity. Super happy to see you finally. Yeah, it was quite of a process, quite of a journey. And I'm super happy that we have the opportunity finally. So, and of course, all egg, all egg if it's, you know, we usually meet each other in person, but you know, virtually it's fine as well. Just less, you know, less food on the table.

Michael Finocchiaro

Absolutely. Yeah. Me too.

Oleg Shilovitsky

Thank

Michael Finocchiaro

Well, I'll come to Brookline and we'll all hang out whenever I can. But from Paris is a little hard. I'll take you guys to some places here in Paris when you come, okay? All right.

Maor Farid

You must, you must.

Oleg Shilovitsky

.

Maor Farid

sounds great, but it's a hell of a competition when it comes to compare Paris to Brooklyn, but but we're up to it. ⁓ So yeah, yeah, in my background, basically, you know, I dreamt to become a mechanical engineer when I was in 10th grade, right? When I saw Sally Edwards for the first time. And then I started Yeah, two years later, I joined some kind of a elite academic program to train mechanical engineering. So I completed my bachelor's, master's and PhD at the Technion, mechanical engineering. Then I started to work as a mechanical engineer in a secret unit that reports to the prime minister's office. And then I realized that I've done the worst mistake of my life, studying mechanical engineering and it's related to the story behind Leo AI. We're going to talk about it in a bit. So I worked as a mechanical engineer, then pivoted into AI and I was an AI researcher.

Oleg Shilovitsky

and and

Maor Farid

In the defense industry, we developed AI algorithms that identify terror targets for homeland security purposes. And then finished my service as a captain and using Fulbright scholarship, went to MIT to pursue my postdoctoral training. Yeah. And also mechanical engineering and AI. And then, you know, I partnered with my best friend and the most brilliant person that I know, to start Leo AI that basically tackles the problem that

Michael Finocchiaro

Nice.

Maor Farid

we experience in first time as mechanical engineers.

Michael Finocchiaro

Awesome. Well, I'm looking forward to learn more about that as we go along. Oleg, you want to talk about Openbomb a little bit?

Oleg Shilovitsky

⁓ Yes, absolutely. So first pleasure to be here. Great to see you. We're getting some echo, Michael, and I don't know where it comes from.

Michael Finocchiaro

I'm not getting in on my side yet. And I have echo cancellation on.

Maor Farid

I hear it too, by the way.

Michael Finocchiaro

All

Oleg Shilovitsky

Yeah, it's kind of strange. anyway, so I think it's not, it disappeared now. So again, Michael, pleasure to be here. Thank you for inviting me. Like Maur said, we usually meet each other in person, different places, but like we do it digitally and virtually now. So I've been around in PDM and PLM development for the last couple of 20 probably, 20.

Michael Finocchiaro

You're better in person. decades.

Oleg Shilovitsky

20 plus years, right? So coming to work in different companies, ⁓ worked for the SOC system. ⁓ So frequently visited Paris. still remember, yes, exactly. And then I worked for Autodesk and in between a few startups, all of them were around development of PDM and PLM technologies.

Michael Finocchiaro

We met there a few times.

Oleg Shilovitsky

So ⁓ my education background was in construction and ⁓ computer aided design. So, and somehow I moved to ⁓ mechanical and moved to manufacturing. And then I found that these two fields, construction and manufacturing are getting close and intertwined today. So that's a very, very, very exciting moment. And I got very passionate about everything that related to data. So my previous company was about how to develop semantic search for engineering data and then OPDM and PLM development. So at OpenBOM, it was a dream how to change the way people are managing data and organizing information. And as we started, then later we discovered all advantages of the AI and graph technologies. So would be happy to speak about this more, but thank you. Thank you for inviting me and great to be here.

Michael Finocchiaro

That's my pleasure. usually I start asking by before I know, Mauro, you just released Leo AI and Oleg, you just had a new release relatively recently. So when you were first starting out on that journey, adding more AI into the product, ⁓ What was your feeling about AI beforehand? Did you think there was a bit of hype, but maybe it was powerful, maybe we'll try to use it? What was the feeling initially before you actually started developing?

Maor Farid

It's a very interesting question and you know, let me be blunt a little bit, know, I started doing AI, let's say like this, before it was cool, before the hype, right? So I really can't really stand people throwing AI, those two letters on everything. And I don't think that people should think about, hey, you know, we need to do AI more. No one really cares. AI is just, you know, a bunch of statistical algorithms, right? And it's not, it should not be a goal by itself. So I really am tired. right, about ⁓ people talking about AI as a hype. And we are focusing on providing real immediate ROI, you know, right, by using AI, but doing things that were not possible before AI. And AI is now natural to solve those kinds of problems. So if we could solve those problems, those business critical, domain specific, and the mechanical engineering problems without AI, we'll do it. We do not care about AI per se. So this is, you know, my two cents and... ⁓ And I'm saying this as an AI researcher, right? And mechanical engineer. So yeah, sorry for being a little bit blunt, but it is what it is.

Michael Finocchiaro

Right. No, it's okay. when you were doing the research, did you immediately see the potential or was it sort of a, or you were a little maybe skeptical at beginning? How did that go? I mean, since you were an AI researcher and a mechanical engineer, did you see the two coming together as quickly as they did?

Maor Farid

I saw the problems even before I knew what mechanic what is right so i'll show you a quick story that will emphasize. How the problems are super. You know, yeah, yes. So makes a lot of sense. So we can talk about ⁓ time to market and different ROI matrices. But I think this story emphasizes it better than everything. So I started to work as a mechanical engineer in a place that it's a dream of every mechanical engineer, right? It's a team that is, it's an organization that develops a very specific strategic weapon, right? I cannot, of course, elaborate about this and their

Michael Finocchiaro

Delated.

Maor Farid

data and knowledge was a treasure in a global scale, right? This is literally a new kind of science, a specific kind of science and engineering. And their PLM was the treasure chest. And you one time we had ⁓ an operation that should take place in six months from from this date. And we had I had to design a new kind of dynamic ⁓ suspension system. But it was not the new kind. It was literally a modification.

Michael Finocchiaro

Hmm. ⁓ new kind of dynamic suspension system, but it was not a new kind, was literally a modification

Oleg Shilovitsky

you

Maor Farid

of a system that exists in 200 or something previous designs. Long story short, I had to spend two months scrolling through the PLM, PDM, trying to find what's different assembly that I should leverage and use and ask dozens of people. No one really remembered. So I needed to reinvent the wheel.

Michael Finocchiaro

of a system that exists in 200 or something previous designs. Long story short, I had to spend two months scrolling through the PLM, CDM, trying to find what system I should use, and then dozens of people, no one would remember, so we did reinvent the wheel.

Oleg Shilovitsky

Thank

Maor Farid

So, and this is the main problem, know, 70 % of the organizations.

Oleg Shilovitsky

you

Maor Farid

are suffering from the fact that their engineers need to reinvent the wheel, redesign stuff that are already in their PLM,

Michael Finocchiaro

Every single time.

Maor Farid

PDM. Right. Not only time, but they, this is a great way to make a lot of design mistakes, not using vetted previous designs and just to tire the system, their peers, et cetera. So it's, this is a classical use case for AI, right? AI is not great in designing the next generation of ⁓ time machines, right? It's not, it's not designing a new kind of science or engineering, but it's awesome on taking your data and just doing interpolations, not extrapolations, if it makes sense. Right. So this is how we do. And, know, by introducing the first AI model that can understand CAD, right? The first thing that this is a prerequisite of what I'm mentioning. So we connected this into

Michael Finocchiaro

taking your data and just doing interpolations, not extrapolations.

Oleg Shilovitsky

.

Maor Farid

organizations PLM system PDM and then we can answer questions find the best fit parts and optimize their designs.

Michael Finocchiaro

So that's cool. That's on the CAD side. So Oleg, you're looking more at the PLM side, right? So did you see the potential of AI when you were divining OpenBOM?

Oleg Shilovitsky

Yeah. You see, this is a very interesting thing because I will tell a story and it's amazingly, I I never heard this more story that he just told, but it's amazingly connected to each other. So I will take you a little bit back and I will say that long time ago, someone told me, you see the SQL database where I invented in order to organize data in a better way. I said, how? And they said, is a language called SQL and once you get everything in SQL, like the problem is solved. Okay. Got it. You see? I mean, that's for real. So, and then basically I look into all the story and I said, well, you know, the problem is the same. People want to organize data. The problem remains the same and the technology are going to change. And I said SQL database are not good to organize data and went to semantic web. So we went to semantic web and developed search algorithms, but it's in essentially it was a graph behind this. So we developed search algorithm, then we move forward.

Michael Finocchiaro

Yeah. Of course.

Oleg Shilovitsky

When I said everything in organization is connected and the of materials is probably one of the most common denominator when people speak about organization of information. got the engineering, manufacturing, service, it's all connected. And how to connect it all together, we need to get a better data management. And that was already after...

Michael Finocchiaro

Like everything. and how to connect all together.

Oleg Shilovitsky

someone said that cloud will solve the problem after SQL and after someone who said social will solve the problem. mean, remember it's all technological things that are coming after another. And we said, we can create a technology online platform that on one side will be global because everyone is connected. All PLM systems are isolated. Like you get 20 PLM systems sometimes in one large company, you get every database. And we said we need to get global systems, like if Google is global, why we cannot be global. And on the other side, we said we need to have a better technology to manage connections, which was a graph database. And by that time when they created OpenBOM, that became a differentiator. So we created OpenBOM platform based on absolutely new data management technologies, including graph database that can scale with amazing capabilities of connection of data. ⁓

Michael Finocchiaro

you

Oleg Shilovitsky

And once we build it, we build what we call a global product knowledge graph, which basically connects all data. And then like Maurer said, AI is statistical algorithm. We just put it on top of this naturally. because your data is the most important element there. And this is how I think it's all pieces started to connect. And we've been talking in the open bomb about intelligence and graph intelligence.

Michael Finocchiaro

And this is how I think it all started to connect. And we've been talking in the open bomb about intelligence and graph intelligence

Oleg Shilovitsky

and ability to find the right ⁓ components and ability to analyze data, now we're just getting another layer of ⁓ technology that can help us to do it. So that's how all these pieces are interconnected. So I'm like, I very much share Maor's opinion that I don't like when someone is trying to come and say, now we have AI.

Michael Finocchiaro

and the ability to find the right opponent and ability to analyze data. So we've been getting another layer of intelligence that can help us to do it. So that's how all these pieces are in the collection. So I very much share my opinion, but I don't like when someone is trying to come and say, now we have AI. All the problems are solved.

Oleg Shilovitsky

I mean, yeah, just put the AI logo on this. Especially I don't like when someone is starting to say now we have something AI native. That's a very interesting piece. mean, especially when something exists already for 10, 15 years, now we have it AI native. We had the same with the true cloud like 15 years ago.

Michael Finocchiaro

Yeah, we had, because we had VMware and mainframes with Charity Cloud. So another topic I wanted to address that I'm sure you're both intimately familiar with is of course we've been talking for, actually I did this last year, 2024. I was already using AI without being a programmer to build stuff, right? And now there's a word for it called Vibe Coding. And it still sounds a bit hype and yet...

Oleg Shilovitsky

Yes, that's...

Michael Finocchiaro

I think there's nobody in their right mind would develop any software without AI today. So my question to you guys is like, is AI ⁓ integrated into the way you're developing Leo AI and OpenBOM? Do you do the scaffolding? Do you do the brainstorming? Is it the test cases? Where is AI? The most useful and where is it super annoying? Like, you know, how AI can often lead you down this path and suddenly it gets stuck and you can't really, it's got to back out again and start over because it can't understand what you want. That's happened to me anyway. ⁓ So I'll throw that to Mauro first. How is a Leo AI actually using AI in the development process?

Maor Farid

So as I mentioned before, the power of AI in general, before we talk about Leo is to get access to data sources. know, ChaiJPT is doing this on the entire internet, right? But they have access to engineering sources and source this information very quickly based on not only, you know, keywords, but on the semantic level and provide real time answers. for the user so what's the parallel for mechanical engineering is imagine that you have an AI that can understand all the data types the modalities of data that you have which means cabin text right basically and it's connected into your PLM data and Windows directories and everything all your data sources dynamically and ever syncing right and then when you ask a question during the design for example what's the best part that I should use here And when you say here, Leo can understand because it's connected to your CAD tool and your PLM. It knows what you mean by here, this whole, this ⁓ surface, et cetera. And it is, we could find demo standardized parts that you are looking for and, and, or documents. If it's a, for example, what's the best coding that I should use here. It can find guidelines from your organization and information from the Bibles of engineering.

Michael Finocchiaro

you are looking for and or documents. it's for example, what's the best coding that I could use here. You can find guidelines from your organization and information from the Bible of Engineering and you can answer questions backed by sources that you find. If you're for parts, you can find parts either from 120 million member parts that we have.

Maor Farid

And can answer questions backed by sources that it finds. And if you're looking for parts, it can find part either from 120 million vendor parts that we have, or from your PLM. And it could tell you what are the most common and standard parts that fit this ⁓ requirements. So this is what we're doing. And maybe one word about vibe coding for whom that doesn't really know what it means.

Michael Finocchiaro

or from your PLM and it can tell you what are the most common and standard parts that fit this requirements. So this is what we're doing. Maybe one word about vibe coding. For those who doesn't really know what it means, I don't really mean anything. It's a way to turn a prompt, a request, into a software product.

Maor Farid

I don't really miss anything, but it's way to turn a prompt, a request into a software product. I'm from the camp that says that it's not going to be available in mechanical engineering in the foreseen future because of a fundamental difference between code and mechanical, right? The mechanical domain. First of all, the tolerance for error, right? It's very, it's relatively low. If you're designing a software that should be, I don't know, a nice mockup for

Oleg Shilovitsky

you

Maor Farid

I don't know, a front end design for a restaurant management. And you have a mistake. So what, so what, okay. If you're doing this for a mechanical thing that should be in the hands of a customer and it could, or a weapon that could explode, not in the right time or a wing of an aircraft, right? It's a major problem. The other thing is, the number of decisions to be made, right? If you're, if I divide.

Michael Finocchiaro

I don't know, a front end design for a restaurant menu. Any other mistakes? So what? So what, okay? If you're doing this for a mechanical thing, that's giving a hand to the customer, or a weapon that would explode not in the right time, or a wing offense, an airframe, right? It's a major problem. The other thing is, decisions to be made, right? can it, you can do by engineering.

Maor Farid

engineering. I don't know, coffee machine. So just imagine how many decisions the engineer, the design, like manually a coffee machine, how many designs did he make or she made? A lot of them. And me as the owner of the product, as the manager of the product, I need to go through all these decisions and vet them. If I get the same output from an AI that design a coffee machine for me, I would need to go through hundreds

Michael Finocchiaro

Just imagine how many students engineer the design, like manually a coffee machine, how many designs did he make or she made? A lot of them. And me, as the owner of the product, as the manager, I make the decisions and vet them. If I get the same output from an AI that design a coffee machine for me. I would need to go through hundreds of decisions that I should check. for us engineers, I see more than three questions, saying, OK, forget about it. I'm going to redesign it myself. this is another thing. But Leo, do not believe in the vision of live coding. I'm sorry if I'm disappointing here.

Maor Farid

Of decisions that I should check. And you know, for us engineers, when I see, when I have more than three questions, I'm saying, okay, forget about it. I'm going to redesign it myself. So this is an, another thing. at Leo, we do not believe in the vision of vibe coding. I'm sorry if I'm disappointing here. ⁓ some folks, some listeners, but we do not believe in a Jarvis kind of vision that, design a coffee machine for me and is going to generate it for you. We believe in.

Michael Finocchiaro

We believe in arming mechanical engineers with AI to make faster decisions and maybe in next months to generate and design smaller centers in a level that engineers could vet and validate, not a full blown product.

Maor Farid

Arming ⁓ mechanical engineers with AI that could make them make ⁓ faster decisions and maybe like in the next months to generate and design small assemblies in a level that engineers could vet and validate, not a full blown product.

Oleg Shilovitsky

and

Michael Finocchiaro

Okay, thank you. How about you, Oleg?

Oleg Shilovitsky

Well, I think it's an interesting question because AI comes to us through the different places. I think when ChatGPT came in 2022, I remember I was sitting with the team and I said, everyone now put a piece of paper on your screen. And then this piece of paper we wrote, before you do something, ask ChatGPT. And that was very useful because... It's about making decisions. So how to make decisions at any moment of time you are, you just need to have something that you can ask and you can get, like Maur said, get the right piece of information. So it came down ⁓ for our development team that they can different tasks that they can do using different AI tools and make them more efficient. We never thought, ⁓ you know, we can write

Michael Finocchiaro

get the right people. So it came down for our development team to make different tasks that they can do with different AI tools to make them more efficient. We never thought more. We can

Oleg Shilovitsky

We can write a prompt that say develop OpenBOM and then go drink coffee and then come back and we get OpenBOM

Michael Finocchiaro

write a Chrome that they develop all the time and then go to the spotter and then go back and develop.

Oleg Shilovitsky

redeveloped in six months like someone suggested. So that's not going to happen. But then it's proliferate from our internal usage also towards how our customers can use OpenBOM. And our customers might decide, like may get some information access more efficient. This is where the conversational intelligence. comes because now we can get answers to particular questions faster. For example, we can say, bring us where I used and export them to Excel, and we don't need to make 25 clicks or five clicks or whatever it takes. So that's just much easier way to do it. Now moving forward, we suggest that we can absorb the knowledge that we have inside of OpenBOM. and also combine it with the knowledge from other systems as well, like for example, the OEI. And we can help people to make specific steps and decisions. Like if they want to create a bill of materials from scratch, we can generate bill of materials for something. I it will not be final. It will not include all decisions, but it will give to someone

Michael Finocchiaro

And we can help people to make decisions. I think they want to create those materials from scratch. We can generate those materials. But it will not be final. It will not include those decisions. But it will give to someone

Oleg Shilovitsky

⁓ starting point. I had an advisor many years ago, he said the hardest part in the development of something is to come with a bad draft. So if you can create a bad draft, then you can improve it. now can think about AI is doing bad draft. Now, build materials, essentially it's a recipe. You want to come with the recipe of your dinner as a bad draft and you can add different pieces. So you can do the same.

Michael Finocchiaro

a starting point. Like I had an advisor many years ago, he said the hardest part in the development of something is to come with the best draft. If you can create a best draft, then you can improve it. you can think about AI is doing best draft. Now, build materials essentially is the recipe. Like you want to come with the recipe of your dinner with the best draft, you can add different pieces. So you can do the same. in the, if you think

Oleg Shilovitsky

If you think about

Michael Finocchiaro

about the materials, because today it's also combined, it's mechanical, or for example, or speaking about mechanical, but also electronic, like, also software, all these other things, so it's combined and also, and give some perspective on this decision. That's how we see it's gonna happen. So it's not like we do everything.

Oleg Shilovitsky

build materials because today it's also combined, it's mechanical, where, for example, more speaking about mechanical design, it's also electronic design, also software, this other thing. So to combine this whole information and give some perspective on this decision, that's how we see it's going to happen. So it's not like we do everything, but we can say, okay, create a draft build materials for your coffee machine.

Michael Finocchiaro

But we can say, create a graph of materials from your coffin machine. You can do a different part. Combine the things. You can absorb some mechanical design. You can absorb some electronic. And you get the final release. And then before final release, as part of the coming features of open-bomb, you'll be validating. So you can go validate bomb. You'll find your 25 errors. And those errors need to be fixed before.

Oleg Shilovitsky

can do a decent part, combine these things, and then you can absorb some mechanical design, can absorb some electronic pieces, you can absorb some software, and you get the final release. And then before final release, part of the coming features of OpenBOM will be validate this BOM. So the feature will validate BOM, will find you 25 errors, and say that those errors need to be fixed before you will release this bill of materials, because mistakes in the BOM when you find them early,

Michael Finocchiaro

you will release the bill of materials because mistakes in the form when you find them early

Oleg Shilovitsky

They cost a little, when you go to production, they cost a lot. So the soonest you can run those queries, so you can think about another way to use OpenBOM. And then we can combine information that comes through people and the different systems and we absorb it, combine it to the knowledge graph. And

Michael Finocchiaro

they cost a little but then you go to production they cost a lot so the students you can run those queries so you can think of open box and then we can combine information that comes through people and the different systems and we absorb it combine it with our graphs and

Oleg Shilovitsky

this is how we get the better answers. So it's all combined together. It's not, I'm very much also in the not... saying that we will get some magic that will say design everything or create everything, but it will be in the small pieces that will give you better decision.

Michael Finocchiaro

So in the current release of your respective software, when the I mean in the case of the AI it's very obvious that you're using an AI because you're in a chat You guys are basically a chat interface right on top of the CAD tool. But ⁓ what other touch points are there between the AI and your application, the AI and the user? Are there also things that you're using AI for in the background that the user doesn't necessarily see? Or is it all this stuff that you're just seeing on the screen, the chat between you and Leo or you and OpenBOMB?

Maor Farid

⁓ got it. So there is a lot behind the scenes. So just, just think about it. Yeah. When the user asks us, asks a question, for example, what's the best material that I should use for this application. And it gets the sources from the Bibles or from his organization and the same for, for, for part reuse and part search. We do have a lot of analysis behind the scenes a lot. And I will explain like.

Michael Finocchiaro

That's what I thought.

Maor Farid

the model that we developed the model that we developed like. You know, Chagy Bt four it's model that called LLM large language model, right? And Claude and Gemini. So we developed what's what we called the world first L M M large mechanical model. So what did it actually do? So it's the, it's the first to not take only words as tokens. What's called tokens, right? Tokens is the

Michael Finocchiaro

Nice. ⁓

Maor Farid

It's words, for example, in the case of LLM and those were trained Claude and Chai GPT to combine them into a sentence that makes sense. So we're doing the mechanical equivalent. We trained Leo, the LLM on millions, hundreds of millions of machine parts as tokens. And it knows how to combine them into an assembly that makes sense in terms of the organizational guidelines, the Bibles, the restrictions, et cetera.

Michael Finocchiaro

⁓ Mm. ⁓

Maor Farid

So when we take this AI model that was trained on the Bibles and we plug it in into the organizational databases, of course, his enterprise grade security, SOC two certified GDPR, et cetera. So we map all their CAD assemblies and parts and text files. And we do process it called indexing, right? Leo understands what's the semantic.

Michael Finocchiaro

When we take this AI model, which was trained on the Bibles, and we plug it into the organizational databases, of course it is enterprise grade security, certified GDPR, et cetera. So we net. all their CAD assemblies and parts and text files. And we do process the code indexing, right? Leo understands what's semantic.

Maor Farid

the semantics of those of those files. And then when the user asks a question, Leo did analysis already, so it can source it very quickly and not go through all this mapping again and again. So, so imagine during the night after you integrate here for the first time, it's going through your data, understand them. And then it's just ⁓ in a heartbeat, it's able to provide you with answers.

Michael Finocchiaro

So is it sort of like a a super engineer looking over your shoulder and saying, hey, you you ought to be doing this.

Maor Farid

Exactly. But the super engineer was going through all the, you know, 1 million Bibles of engineering and read all your documents. Yeah, this is it.

Michael Finocchiaro

Bye bye. Very cool. How about an open bomb? What does it look like for the user of open bomb?

Oleg Shilovitsky

Yes, there are multiple places. So one of them is very simple. It's chat because in chat we use enlarged language models and then we bring data in context through OpenBOM MCP server. So this is the simplest element because OpenBOM provides tools that then can be combined in a chat using large language models. But there are more sophisticated and more ⁓ advanced use cases that are coming from rag, I'm ⁓ sure you know retrieval, augmented generation. So then it was another way to provide more precise answers, more precise results from large language model, which was a graph rag, which used some graph algorithms for this. Now we're taking it forward and making it more specific. And I'm not sure the name will go public like a bomb rag. or whatever else, but yeah, we certainly use it inside because this is another way to improve the precision of data because like GraphRag, use graph algorithms to bring the right data in the context of large language models. If you think about BOM, it has a very specific data structures, like the data that knows that there is an engineering, manufacturing, maintenance and sales BOM.

Michael Finocchiaro

has very specific data structure. The data is known that there is an ADDX when you're actually in the sales box.

Oleg Shilovitsky

So if you start providing this specific information, when someone is asking you a question about parts that used by current customers, we can navigate to the right places of data that are expressed in the maintenance bomb.

Michael Finocchiaro

to start providing this specific information, when someone is asking you a question about parts that used by current customers, we can navigate to the right places of data that are spread in the maintenance box.

Oleg Shilovitsky

So without that, this data is not marked. So this is how we can bring more specific data and lead to more specific... correct answers, precise answers. So that's how it can be used. And it's also how I see the future integrations of technologies like LEO and OpenBOM, because ⁓ it's about how to make data available and make data available to make right decisions. like Maor was speaking about millions of parts available and the engineering knowledge. Now we bring specific data that is managed by OpenBOM.

Michael Finocchiaro

That's it can be used and it's also how I see the future integration of technologies like Leo and Opel. It's about how to make data available and make data available to make right decisions. I was thinking about millions of files available in the engineering college. Now we bring specific data that is managed by our department.

Oleg Shilovitsky

And that's improved the precision of the same recommendations that

Michael Finocchiaro

And that is the precision of the name recommendation.

Oleg Shilovitsky

the engineer wants to get.

Michael Finocchiaro

There's actually a question on the chat whether Leo actually, you guys created your own foundational model, your own, so the LMM is actually your own foundational model.

Maor Farid

Mm-hmm. Exactly.

Michael Finocchiaro

Very cool. ⁓

Maor Farid

And, you know, a foundational model is, is this, it's basically an enabler, right? It's an edge that will basically can tackle. Dozens of use cases. And we're super happy not only to serve our customers, right? But also to serve all eggs and open bombs customers. I mean, as, our, our esteemed partners and also like other place, like on chip, our esteemed partners. And when the LM is in the hands and Leo is in the hands of very strong teams like open bombs and on shapes that they know and obsessed about their customers needs and their data structure. And basically they can apply to solve many, many use cases like, like open bomb is, is, is working on right now. And just, you know, it's not another AI that can ⁓ answer nice questions or do like fairy dust magic.

Michael Finocchiaro

needs and their data structure and basically they can apply it to so many, many use cases like OpenBombs is working on right now. It's not another AI that can answer nice questions or do fairy tests.

Maor Farid

But to tackle real problems that, and all that he knows, he's obsessed about those problems for a long time. So I think, know, OpenBOMB's customers are super lucky.

Michael Finocchiaro

Mm. ⁓ Maybe the, well, I didn't really pose the question this way, but there's also sort of different levels of AI, right? You've got the, I'm gonna wrap a chat bot, ⁓ the chat bot, the chat GPT wrapper, which is just that sitting on top of documentation thing. But then you've also got agentic AI, right? Because thanks to Anthropic and MCP, now you've got AI's talking to each other. So potentially you can have multiple, I suppose even Leo could have an agent that really understands chamfers, another one that understands milling, another one that understands assemblies. Then you could have these specialized piece of knowledge. I suppose one of the fears or one of the problems you run into is the lack of determinism. There's too much probabilistic stuff and then you end up with hallucination. I'm thinking if you focus your agent really down on one specific problem,

Maor Farid

Mm-hmm.

Michael Finocchiaro

He has less things he can hallucinate about because he only knows about that one domain. He can't just go off and give you photographs of Sabrina, a carpenter or something instead of what you're actually asking for. So is that how you're doing it with Leo also?

Maor Farid

Yes, so similar. So let me say a few words about hallucinations and why they're absolutely horrible and for mechanical engineers. And I will start with a story. know, when we just started, when we started on May, 2023, before writing a single line of code, Moti, my dear partner and I, we started from serving and interviewing more than 900 mechanical engineers.

Michael Finocchiaro

Nice.

Maor Farid

And we itemize the most time consuming tests that are facing today. But also, you know, at end of the conversation, you know, after the second, second glass of beer, what's problem with chat GPT really? Like, why, why you hate it so much? Because they were radiating hate towards chat GPT. And they told me, look, more. I prefer Google 10 times more than chat GPT. Why? Because Google never lies. Right. It retrieves shiggly.

Michael Finocchiaro

All right.

Maor Farid

or machinery handbook in the fifth page, but if he doesn't know an answer, he doesn't show an answer. Chagypt tells you, ⁓ yeah, yeah. And Chagypt, if you're asking what's the best bolt I should use, he's gonna tell you, you gotta please use M8 bolt. And then you're asking, of course, of course, smartest prettiest, I mean, what pretty eyes you have, Such a, yeah.

Michael Finocchiaro

Yeah, zero search results. And what a brilliant question. And what a brilliant question. You're the smartest guy in the universe. Yeah. ⁓ Yeah.

Maor Farid

So, so he tells you with a high confidence, you should use M8 port. And then you're asking, are you sure? And then he tells you, you're right. It's M12. I cannot beat an airplane based on this, based on this, yeah, you're right.

Michael Finocchiaro

And then you ask again and it says M9 and then you ask again it says M8 and then it goes back around again. ⁓

Maor Farid

Again, sorry, exactly. what basically we, the first problem that we needed to solve really is not understanding CAD, is the trust issue. So we needed to provide an AI that an engineer could feel.

Michael Finocchiaro

Exactly.

Maor Farid

100 % sure that he can trust or she can trust the answer. So what Leo basically does is ⁓ differently than chat GPT. First of all, it was trained on not on Reddit, by the way, chat GPT is hallucinating because for us, mechanical engineers, 46 % of the time, not official data. This is our survey. 46 % of the time when you're asking a mechanical related question, it's going to be an error. Why? And this is a real data. 40 % of the underlying data sources under the hood of Chai GPT is Reddit. And Reddit is exactly the opposite of how us humans, how we build engineering, right? It was an art going from a father to son. Now from Milt Standart and Izzo to an engineer. Reddit is the opposite. It's anonymous people writing random, sorry, ⁓ shit online. Yeah.

Michael Finocchiaro

Yep. Yeah. with no fact checking. Yeah, with no fact checking.

Maor Farid

Exactly. So, so first of all, we trained Leo to answer questions only based on the Bibles or your sources. If it doesn't know the answer, tells you, don't know. If it knows the answer, it gives you a source that when you click on it, it opens up the right paragraph in the right page in the right book, saving you the scrolling time. So you immediately know why the decision was made like this. Of course it understands CAD, unlike JGPD that understands well only language is a language model and integrates into the PLM that JGPT cannot do, especially now. So this is the differentiation regarding the, the new wave of agentic AI in other terms for the layman is basically an AI that does automation for you. So imagine that you could tell, Hey, Leo, ⁓ design, I don't know, do the fillets on the entire, entire assembly or whatnot. So we are working.

Michael Finocchiaro

of the entire assembly or whatnot. So we are working on this. It's gonna be ready soon. But always keeping in mind that engineers would love to have a very small automation type of it. And we're very surprised how small it is. Right to the fillet at this kind of hole. not the product level. Very, very good. That's sort of the way I was thinking about it too.

Maor Farid

on this, it's going to be ready soon. But ⁓ always keeping in mind that engineers would love to have a very small automation space for them. And we were very surprised how small it is. ⁓ Right? Do the fillet at this kind of balls, but not in a big product level, right?

Michael Finocchiaro

How about with OpenBOM? Is that you have the same impression or you? I would think PLM lends itself to having multiple agents because we have so many disciplines inside of PLM between change management and configuration management and CAD management. So what do you think, Oleg?

Oleg Shilovitsky

Well, there are two aspects that I want to bring here at one. I really like the trust element of the data that you can trust because this is where the first value of what we develop comes into. ⁓ We know exactly the data about the product, what components are used in different products and different assemblies and different subassemblies, how they're connected to other elements. So when someone wants to know what is the best screw to use for particular design, it automatically can bring a data that related to all designs that you've been using before. this is, know, when engineer can't find something, they create a new one. So, and then you get many parts in the company. So, so in OpenBOM use case, because we have the product knowledge graph that related to specific company, That's very easy for us to help to navigate to the right data, to trusted data that, for example, already used in the product. And that's the foundational piece. And then we take the next level. And the next level is open-bound platform. Now open-bound platform is global. And many companies are using this. So this is our vision. It's not something that is actually happening now, but we can use different statistics. about different companies using different data. And if those companies will allow to us to use these statistics to bring this data to someone else. If you think about OpenBOM Global Platform, like Amazon with many people buying something, you can bring the same knowledge in the context of many companies developing many products using many components. So we can say that, for example, you're selecting part that by the way failed in 70 % of use cases and other products by someone else based on the information that we have. So it's a foundation to create layers of data. And that's very valuable because like Moor said, it's like you don't want to build your product based on Reddit knowledge, right? You want to build your product based on the precise knowledge of companies that build particular products. And this is where OpenBoom brings this knowledge.

Michael Finocchiaro

So it's a foundation to create layers of data. And that's very valuable because you don't want to build your product based on Reddit knowledge. You want to a product based on the of companies that build particular products. And by

Oleg Shilovitsky

And by integrating those things together, you can start getting information.

Michael Finocchiaro

integrating those things together, you can start getting the foundation.

Oleg Shilovitsky

So that's about trust and data. And the second you ask the question about agentics, I think there is a potential here to get particular tasks exposed that can be used for ⁓ different scenarios. Like if OpenBOM provides MCP server, This MCP server and tools can be used in different scenarios. Like if someone is developing change management process or validation or supplier qualification. So those scenarios can be programmable and used by even working particular tools. So that's how I see the immediate use will be. And again, invoking those tools can be used for different purposes. For example, you can invoke tools

Michael Finocchiaro

So those scenarios can be programmable and used by invoking particular tools. So that's how I see the immediate use will be. And again, invoking those tools can be used for different purposes. For example, you can invoke tools

Oleg Shilovitsky

and make cost estimation.

Michael Finocchiaro

to make books.

Oleg Shilovitsky

So you can make cost estimation based on different data. You can make cost estimation based on the global data. You can make cost estimation based on the particular catalog that Leo AI has an access to. it's like when, yes, when the species starting to connect together, this is where this agentic to me becomes a real thing and we can use it. So it's not necessarily about running particular process.

Michael Finocchiaro

The McMaster and stuff.

Oleg Shilovitsky

but it's more about getting intelligence. So don't think MCP will replace APIs for communication because when I need to send something to ERP system, I still can use ERP API much more efficiently. I don't need MCP server of ERP system. That's unlikely will be a good idea because OpenBOM can use, for example, REST API for ERP system. But when we want to create some intelligence on top, This is where we can start getting information from different MCP services, bringing them together. So that's the agent that, for example, will make an assessment of the bomb validation. So it's making multiple agents to validate these bombs, and that's becoming like a first level validation, second level validation, third level validation that includes particular tools that can be used.

Michael Finocchiaro

This is where we can start giving information from different communities together. For example, we will make an assessment of the bomb validation. Making multiple agents to validate the bomb, it's becoming like a first-person validation, second-person validation, third-person validation. This includes particular tools. That being said, think that the next generation of digital thread will be agentic in the sense that we shouldn't need to have to write these integrations anymore because the MCP should be able to figure out the input and the output and the right protocol to use, right? I mean, that's sort of the philosophy of just leaving the backend, the plumbing to the AI as long as there's some human oversight, right? ⁓ But that's super interesting. like, before we go to the customer side in like five minutes, Over this journey that you've had, ⁓ and you've both put out product that's very, that has AI more at the center. I'm not going to say AI native because I'll get killed by Alex. So I'll just say AI more at the center of the whole concept. And I think a Leo AI even has a letters AI associated with it. But what have you learned about AI over? Well, let's just even say over the last two years, when we've seen everything change every day, almost every day, right. Including.

Maor Farid

Hahaha. Mm-hmm.

Michael Finocchiaro

⁓ vendors saying, actually we lied and we're actually training on your prompts after we told you we weren't. you know, a lot of changes. how, what have you learned and what are the, the, some of the takeaways for the audience about, know, what, what's real and what's hype. Because one of the other, of course, a couple of people in the thing were saying, are we, where are we on the hype cycle with this whole thing? So. ⁓

Maor Farid

wow. we've only if you'd like to start from all like it's gonna be even more interesting. So we learned a lot ⁓ more than we learn about like AI because we live we are fully immersed in this technology as an as a researchers, not only mechanical engineers. We've learned about mechanical engineers holding AI, which was amazing. It was eye opening. At the beginning, we thought

Oleg Shilovitsky

Hmm?

Michael Finocchiaro

You could. every day.

Maor Farid

just if I'm going to summarize this, you know, two years and in a few minutes, first of all, we thought they want a vibe engineer. We thought they want vibe engineering. And the market just slept us in slept us with brute force. You know, our heads were just throwing more from Brooklyn to Newton. So and we realized that they want to get rid of the

Michael Finocchiaro

Mmm. Hahaha!

Maor Farid

The boring part, but not the too boring part as well. So we overshoot it into the, know, from people that talking about ⁓ vibe engineering to people that are talking about. Automations, CAD automations, the fillets kind of stuff.

Michael Finocchiaro

Right. The fillets, the chamfers, things like that.

Maor Farid

Right, they want us to reduce clicks for them. But also we realized after a few questions that it just, even though it's annoying thing to do all those clicks, they're not really business critical. No one really cares about them in the strategic level. Right? So the, the, the problems to focus at the, the, the maximum ROI are the knowledge problems, those problems that mechanical engineers won't mentioned.

Michael Finocchiaro

Mmm.

Maor Farid

once mentioned in conversations, but when you're digging deeper, what they really care about what really keeps them ⁓ past the due dates, the deadlines to pay much more for the product life cycle. This, this is the knowledge problem. Effective engineering knowledge in 2025 is still siloed. The PLM is still a locked vault. The minds of engineers is a locked vault. The Bibles are locked behind search and paywalls. And the CAD is not something searchable really. So this what we realized. The boring parts, right? Those are, those are the most.

Michael Finocchiaro

behind search and paywalls and the cat is not something searchable. So this is what we realized. The boring part. ⁓ Fascinating. I love that. What about you, Oleg?

Oleg Shilovitsky

Yeah, you know, let me bring three stories here. So first is about hype. So let's just explain what is the level of the hype. So you speak to customers and we do it a lot. So no one is really asking something specific just by saying, do you have AI to solve and then the name of the problem? Like it doesn't matter. It's like basically everyone believes that it will be an AI that will solve their problem. And this is how they articulate it. Like, I'm making mistakes. Do you have an AI that will help me to not to make mistakes? Great. Excellent. So that's a level of hype. I think it's, you know, as we move forward, it will be over because people are digging deeper in and trying to understand it a little bit more and the knowledge is coming. And then, you know, when we really come to the value, I like to classify everything, like everything that company buys is because of three reasons. It's like either they want to earn more money or they want to save cost, they want to mitigate risks. Like everything that you want to sell, those are three things. you try it. Like I tried, like I came to conclusion, those are only three use cases. So now when we come to AI and open bomb, this is very simple. Like get more...

Michael Finocchiaro

think everything that company buys is because of freedom. Either they want to earn more money or they want to save costs, they want to mitigate it. Everything that you want to sell, those are three things. Try it. I came to conclusion that there are only three of them. Now, when it comes to AI and open home, it is very difficult.

Oleg Shilovitsky

get more, like earn more, that people want to create more orders. So we have customers that create orders using OpenBOM. So if AI will help them to create better orders faster, today they're getting data from CAD systems, generate purchase order in OpenBOM. If they can make it faster, amazing, right? So they will be making more orders. So if someone will be sending mistakes in BOMs and we can find those mistakes early, so we can mitigate risks. and not to use components that will be eliminated next year or next six months. The AI will give a better way to solve this problem. For example, we want to save cost. If we can find a way how to

Michael Finocchiaro

⁓ So if we can find a way how to optimize the design. ⁓

Oleg Shilovitsky

optimize the design, how to optimize build materials. So it's all about particular decisions that we are taking and we see them a lot. Like for example, today on OpenBOMB blog, there is an article with a sample of five companies. everything, like you can go and see the article. Different companies, different things that they develop. There is one common thing between them. The data was chaotic.

Michael Finocchiaro

All right. ⁓

Oleg Shilovitsky

in different spreadsheets, in different places, OpenBOM helped to organize this information. Now they can make decisions. Now, put AI on this, we can make even better decisions. That's the story, which brings me to the third, the most fundamental thing. The data is the central part of everything that we do. So we will get right data. We will, like Morissette, we will be less hallucinating. We will be trusting what we do.

Michael Finocchiaro

you All right.

Oleg Shilovitsky

and we will optimize companies' decisions to get more money, to mitigate risks and save costs.

Michael Finocchiaro

So kind of doubling down on what you just said, both of you obviously have sold into companies of various sizes and ⁓ I usually think of digital maturity on a scale of one to five, one being still Excel and email. five being agentic adaptive digital twins, which is basically the la la land of engineering, right? So I'm imagining that most of the customers you run into are between one and three, probably between one and two or between two and three, but maybe closer to two. And then you bring this awesome AI powered solution. And as Oleg said, garbage, get garbage out. If the data is not great, then the results aren't gonna be great. So when you bring this tool and people realize, my God, knew what I was doing, my data was better, I get more out of the tools. Is there an aha moment? Is it just a slow ripple where they're like, okay, well, we'll get to that a couple of years from now, but we're too busy with, you know, the hitting our deadlines and getting the immediate problem out of the way. there, there's the fact that your, the customer implements one of the, one of your solutions is it create a bit of a bigger wave of changing the organization and fixing these data problems that are plaguing everybody, right? Because of the siloed organizations in the siloed disciplines. Is this a helping thing or is it still an impediment?

Maor Farid

So maybe I'll enter with a story, right? After I met one of my mentors and a brilliant person, Bertrand Sicko, the former CEO of SolidWorks, that now is not only our esteemed advisor, but also an investor in Leo. He said after he saw Leo, he felt the same feeling that he felt 30 years ago when he saw SolidWorks for the first time. And I think... It resonates with a lot of stories that I've heard from those days from the founders of this company. And so there we're talking about 1995 when they started to bring SolidWorks into market. And some of their customers told them, I don't need this. don't need to use it in CAD. You know, I use a pencil and an A0.

Michael Finocchiaro

And so they were talking about 1995 when they started to bring college work to the market. And some of their customers told them, I don't need this. don't need to use it in Canada. Plus Windows, are you kidding? Windows is never going to be able do that, right? It was Unix, but.

Maor Farid

Of course. Yeah, not not Windows, not Unix. I love my pencil and my paper. And then engineering was divided to two, the old guard and avant garde, right?

Michael Finocchiaro

Yeah.

Maor Farid

You know what really happened, right? The old guard doesn't exist anymore. they basically, because the new guard had 10x them. And it's very similar to what happens with Leo today, right? It's a domain specific AI and our customers are seeing 30 % reduction of design mistakes, 35 % of increase in their part reuse. And so it just tells you how business critical it is. But we, in our case, because of the

Michael Finocchiaro

And it's very similar to what happened with Leo. increase in the part reuse.

Maor Farid

maybe formal about AI, our dynamics is different. Engineering leaders reach out to us. They're already convinced and know what they can earn from AI and what Leo can provide them. And basically we're just checking the boxes of security and trust and everything when they see the SOC 2 certificate and everything. It's basically integrating into their systems. and providing them with our services, pretty given. So this is the dynamics that we see right now. it's so to your question, it's not about their AI maturity. Those are people that are looking for competitive advantage. Those are engineering leaders, what we call the, you know, engineering managers, directors, VP, etc. That are looking for the next big thing to 10x them and Provide them with a competitive advantage because the way I is not the best is the only available for mechanical engineering and what we do so. We do we do not face really.

Michael Finocchiaro

So in your case, you're sort of preaching to the choir. Like people are already convinced. ⁓ but. And then. Yeah. So, Oleg, I think in the PLM world, it's not exactly like that, right? It's a little bit different.

Maor Farid

They preaching to us at the beginning what they need and then we preach back. Yeah. Yeah.

Oleg Shilovitsky

⁓ actually, you will be surprised. Let me give you a story. ⁓ Today it's way of stories. It's another story. I like more basically took the word that I've been thinking about and to say organizations are coming after 10x. So, if you have something that is a little bit better, it's like try to...

Michael Finocchiaro

It's a story, another story. ⁓ A little fireside chat. Okay. ⁓

Oleg Shilovitsky

not to discourage people from deciding for the SOPTC or Siemens, but those tools are approximately the same. They might be a little bit better in one and they might be a little bit better in something else. You might like one company and dislike another one. It also happens with all humans. And this is where everyone is speaking now about people. because it looks like people is the competitive advantage. If you make someone to like the saw more than Siemens, you win the deal, right? So that's, again, it's just perspective, okay? Now, coming to 10X is important because SolidWorks was 10X. Because SolidWorks, when it shows up, it was 10 times cheaper and was able to do something that no one else was able to do.

Michael Finocchiaro

to then end. It's huge. and was able to do something that no one else could Well, was the same revolution that Parametric had done 10 years before with ProE, right? It was the same explosion of pro... Yeah, with the cost, lower cost. Yeah.

Oleg Shilovitsky

Yeah, but the cost, but the cost, right? Okay, so here's where I am coming up. So OpenBOM is the only one that you can go online and create your account for 30 seconds. Try to do it with someone else. Try it, okay? And then try to import data to this account ⁓ and start using it. Yeah, it's 14 day trial, but you can collect, we collected about 60,000 companies that registered for OpenBOM to do.

Michael Finocchiaro

Okay. ⁓

Oleg Shilovitsky

something which is the indication that there is a huge underserved space in this. So when the SOA Siemens and PTC are fighting for a thousand to 10,000 customers they have in the world, because everyone already have at least two, three PLMs. So they're trying to outsell or resell or upsell. Agile is discontinued now. Let's sell to Agile customers. I mean, this is the very nice group of customers. love them.

Michael Finocchiaro

Interesting. They're trying to outsell or resell or upsell. Agile is 15 years now. Let's sell to Agile companies. I mean, this is a very nice group of companies.

Oleg Shilovitsky

great companies, amazing companies. The only problem that there is a 90 % of others in manufacturing that don't even can come even close to this PLM system. So it's a lead club. It's great. Absolutely important. It's a complex

Michael Finocchiaro

Great companies. Amazing companies. The only problem is that there is a 90 % of others in manufacturing that don't even close to the PLS. So it's a lead club. It's great. It's really important.

Oleg Shilovitsky

use cases, absolutely amazing technologies. But then we find all these companies that have similar problems, much smaller. They come into us, they interested because they are looking for 10x. So they are looking for technology that can be used and I would encourage you to send me a name of the PLM company that you can go on their website and create an account. You can create openbomb account. In 30 seconds after this meeting.

Michael Finocchiaro

Okay So

Maor Farid

Hmm

Michael Finocchiaro

Okay, so we're almost at the end. thinking, I think a really cool question to end would be since you guys are both, ⁓ you know, some people with very large domain knowledge and there's a lot of people that might listen to this that are worried about AI and worried that, you know, they might be out of a job because of AI. ⁓ What kind of advice would you give it being someone that's a practitioner, especially in your case, Mauro, because you've been doing this your whole career? What kinds of things should they focus on in order to stay relevant in this new age of AI as an engineer?

Maor Farid

So those are two questions, right? If they should scare about AI. I can talk on me on our behalf, as you the first AI, ⁓ enterprise ready AI that is available. So even in our long-term roadmap, we do not have any feature that is going to replace a mechanical engineer. It's not even, it's exactly the opposite of what of our engineering religion. Our main goal, only goal is to make engineering more productive and to make the engineering design more, not only with significant less risk and much faster time to market and much higher part reuse, but also joyful for the mechanical engineer. So this is the first thing. yeah, ⁓ my advice was just to try it out, like just to do your research. Forget about Leo AI. Try to search ⁓ for another AI tool. If you find, just let me know. I don't know. I live in this market for long time. I don't know anyone in production. Find yours, hate it, love it, whatever. But just keep your fingers on the pulse and don't become obsolete because everyone knows that AI is going to revolutionize our domain. Don't stay behind.

Michael Finocchiaro

me know. don't know. I live in this market for a long time. don't know. Thank you. And Oleg?

Oleg Shilovitsky

Yes. And in our vision, we have two things. First of all, we want to help people not to do tasks that they don't need to do. Like no one should combine Excel's manually. It's just a stupid task. but believe me, believe me, there is in every company, there is a chief Excel officer who does it. So, and then the second is that we want to make people ⁓ smarter. So we want to bring intelligence to people.

Michael Finocchiaro

Yeah.

Oleg Shilovitsky

And I, you know, we disclose of the number of years that I'm in the industry. can say, remember the time when Google became widely, widely popular. And back in that time, people said, if I can get access to Google in the meeting, I can be much smarter in the meeting. Okay. Because I will get some answers that no one will have immediately. So, and then I think it's about what AI is becoming as a tool for people to be smarter.

Michael Finocchiaro

I can get access to Google in the meeting. I can be much smarter in the meeting. what AI is becoming is a tool for people to be smarter.

Oleg Shilovitsky

And in order for people to be smarter is to give them right data that will help them to become smarter in this

Michael Finocchiaro

And in order for people to be smarter, it will give them right data that will help them to become smarter. Oleg Shilovitsky (1:00:04) place. Because if I will ask what is the right screw, I mean, great, but I don't need this best screw. Like I need to know for my particular project. So if I give the information about my data and my project, ask for the right screw and that we sell in particular country, then I can get the right answer. So it's about how to be more intelligent. Michael Finocchiaro (1:00:25) That's awesome and I really appreciate you guys taking the time to talk to me today. I think I learned a lot. Hope you guys had a good time too. Maor Farid (1:00:34) Absolutely. Oleg Shilovitsky (1:00:34) Amazing. Michael, thank you for inviting. Michael Finocchiaro (1:00:35) feel like we have even more to talk about. Maybe we can all get together again and talk some more. That was really exciting. Thank you very much. ⁓ Thank you for the audience and have a great afternoon for you guys. So for me, it's the evening already and we'll see you on the next podcast. Thank you. Maor Farid (1:00:50) Hahaha. Oleg Shilovitsky (1:00:51) Thank you very much. Thank you,