Pausing a product roadmap for an entire month to point 700 engineers at a single goal is a significant structural shift, but it transformed monday.com. Andrew sits down with VP of R&D Sergei Liakhovetsky to uncover how fixing core infrastructure and adopting a cell-based architecture paved the way for platform scale. Sergei details the exact framework his leadership team used during their 30-day pause to launch user solutions while maintaining a strict zero-bureaucracy policy. The conversation also explores the new realities of reliability as platforms transition from being CPU-bound to heavily GPU-bound under the weight of automated agents.
Show Notes
- monday magic: A tool for generating initial work solutions and boards using simple prompts.
- monday vibe: An app builder that allows users to create custom applications on top of the monday.com platform.
- Sidekick: The horizontal AI assistant/copilot that works across the entire platform to help with tasks like data management and content generation.
- Agent Factory: A platform for building vertical, specialized agents that can handle specific workflows and roles.
- Connect with Sergei Liakhovetsky on LinkedIn
Transcript
(Disclaimer: may contain unintentionally confusing, inaccurate and/or amusing transcription errors)
[00:00:00] Andrew Zigler: Today, I'm thrilled to welcome our guest, Sergei Liakhovetsky the VP of R&D at monday.com. Sergei, welcome to the show.
[00:00:09] Sergei Liakhovetsky: Nice to be here. Thank you for hosting me.
[00:00:12] Andrew Zigler: Of course we're really excited to have you here and today we're talking about how Sergei didn't just roll out AI to his team. Instead, he paused the rest of the roadmap for 30 days and pointed a 700 person technologist organization at AI enablement as the goal. And they came out the other side with something pretty rare. Every developer using AI daily new platform capabilities, already seeing adoption numbers that most teams are not hitting. In the numbers, they do speak for themselves. We're talking about Monday Magic with 5,000 solutions built in under three months and Monday Vibe with 40,000 apps created in two months.
[00:00:49] Andrew Zigler: Sidekick 150,000 interactions in less than a quarter, and under the hood, you're talking about an insane acceleration of one of their biggest tech debt problems, cutting a [00:01:00] 33 Your investment down to just five months of work and handling time for complex CU customer tissues dropped from three days to one test coverage do doubled onboarding speed sped up to 21%.
[00:01:11] Andrew Zigler: We're talking so many gains. So how do you build for humans and machines at the same time in this world? And what happens when your organization hits AI escape velocity? That's what we're gonna find out today on Dev Interrupted.
[00:01:26] Andrew Zigler: So Sergei, I wanna start by talking about the numbers we just ran through. I talked about all the thousands of apps created by your users, the amount of tech depth that you've eliminated, and every developer using AI daily, and none of that happens without a lot of trust and reliability underneath. And when we talked initially, before this call you, you said something really stuck with me about trust being the currency for all of this, and this is the foundation that lets you ship all of these amazing features for your users. So when you look at Monday.com's journey [00:02:00] from B2C to enterprise scale, what were the first cracks that showed that told you, oh, we need to rebuild this foundation to get ready for this new era.
[00:02:09] Sergei Liakhovetsky: Yeah, so it's a great question, Andrew, and I think that we need to start first all from the, uh, culture of Monday. Monday is a great company where we're focusing a lot on the customer experience and when we're focusing on customer experience. We are looking at, uh, how actually our users will use the system, and this is was the main driver of what whatever we did so far.
[00:02:37] Sergei Liakhovetsky: So first of all, the ux, first of all the experience and after it will look on how the system should work. So at some point when we continued working upmarket, actually we saw that the system, it's lagging behind from performance perspective and from scale perspective. And this is where we started looking [00:03:00] on the different solutions, what we can do in a different way.
[00:03:03] Sergei Liakhovetsky: And this is how the mdb, the first version of MDB right now. We already, in the third version of MDB raised, we started looking on how we're building absolutely different way. How we're dealing with data. So from one side we had the trade off. Okay. Because when you want to have a great performance, you need to think how the user experience will look like.
[00:03:29] Sergei Liakhovetsky: And this is where we decided that we initiate this project of MondayDB and this project ran for two and a half years till we had the first release. Actually, we replaced the entire underlying technology with the data management system that we built. We are using different foundations. We started from SQL, moved to Cassandra.
[00:03:52] Sergei Liakhovetsky: Now we're using also in cache databases like DDB and others and [00:04:00] absolutely different. So we started when. Uh, the, uh, system was supposed to, uh, deal with the boards, boards, you know, in mandates like tables.
[00:04:09] Andrew Zigler: Right.
[00:04:10] Sergei Liakhovetsky: That is keeping a. Thousands of items, and now we can maintain millions of items in one port, in one entity.
[00:04:18] Sergei Liakhovetsky: And this is huge. So first of all, this was the first driver of moving forward with the foundation to support our customers, to gain that trust, especially when we're talking about enterprises, about large customers that are looking for predictive solutions. They're looking for the solution that, uh, will work with greater reliability, with the great availability.
[00:04:44] Sergei Liakhovetsky: And this is what we did the first and the second one that we're doing right now is the cell architecture. In the cell architecture, we are focusing on reducing the blast radius, how we're going to reduce the blast radius of incidents. So in a way that [00:05:00] one noisy count will not take the entire system down.
[00:05:04] Sergei Liakhovetsky: All that together gives a lot of trust to our current, to our customers, and this is what we're pushing ahead because for us, customer experience is trustworthy customer experience.
[00:05:19] Andrew Zigler: So it starts at the beginning by understanding that you had to fundamentally improve your underlying infrastructure and architecture to handle the scale and the complexity needed by those customers. And that's nothing glamorous. That's like getting into plumbing and fixing things and ripping things out, and making it better performance wise with databases.
[00:05:38] Andrew Zigler: This is before any kind of glamorous ai. Work that
[00:05:41] Sergei Liakhovetsky: Yes, yes, absolutely. So if you are looking on the Monday journey, I think it's a bit different journey from, uh, other, other companies, other startups. First of all, Monday started from user experience. And after it moved to deal with performance and scale for [00:06:00] larger customers and after it, we moved to ai. So we are always looking for how to improve the experience for our customers.
[00:06:09] Andrew Zigler: How do you build this culture where investing in those foundations to make that great customer experience is seen as like a first class thing to go after, and it's not just slowing you down. How do you align the culture around that?
[00:06:22] Sergei Liakhovetsky: Yeah, look, when we're looking on the foundations, first of all, we're looking at like, uh, for me, foundation is equal to standards. So if we are going to, uh, our bigger customers and we want to gain the trust and we talk, the trust is the currency for our customers and trustworthy experience, we need to make sure that whatever we're doing, whatever we're building.
[00:06:47] Sergei Liakhovetsky: We're building like a highways, okay, and highways. To build a highways. It takes time, but in many cases, when you're building this highway, you have unlimited speed later on. If [00:07:00] this is exactly our approach here, when we're going to provide the experience and where we need scale and where we need performance.
[00:07:10] Sergei Liakhovetsky: We are building those highways. We're building the robust foundation. We're using cell architecture. We're building our foundations, our resilient data layers, our additional shielding tools that are shielding and preventing incidents and providing resilient behavior. But, again, it's not always the case.
[00:07:30] Sergei Liakhovetsky: For example, right now when we're talking about ai, it's not enough. We need to know how to balance between robust foundations and moving fast. So we are working as the builders all together to identify where we need to move faster, where we need less resilience and less scale, but we need to have speed and we're building those products.
[00:07:54] Sergei Liakhovetsky: So experimenting. From one side and on the other hand, [00:08:00] to have the robust foundation for the solutions that should run in scale and should run with the high performance.
[00:08:07] Andrew Zigler: That's a really sharp observation and something we're gonna talk about more a little bit later too. The idea that, you know, you, you, you have to get the foundations together in order to go fast, and a highway is a promise for speed in the future. That's what's a cool thing about a highway. You can't build a highway in secret.
[00:08:22] Andrew Zigler: Everyone knows the highway's coming and you're gonna go fast in the future. So in the same way you get everyone aligned around the same goal, we're gonna make our foundation amazing. So it sings. That way we can go really fast because we are in this like difficult environment. Engineering leaders every day are having to make tough decisions between do I double down on our infrastructure and spend more time fixing our technical debt, or do we just innovate in speed at the speed of light and just hope that we just catch something that yanks us forward?
[00:08:48] Andrew Zigler: You know? So it's like a real. It's a real balance, I think, and in this world, there's lots of metrics you can use to navigate the success of fixing your underlying foundations and [00:09:00] uptime and data integrity. Those are just, you know, those are pretty straightforward. But what other signals do you or monday.com look for to understand how trust is increasing or eroding with your customers?
[00:09:11] Sergei Liakhovetsky: Oh, it's a, it's a lot of different metrics. It's not only about uptime, it's uh, also about performance. I think that in the new world, especially with AI when we're looking for, our customers basically are looking for a different, uh, latency. And they are, we're looking for instant responses from the system,
[00:09:32] Sergei Liakhovetsky: performance is like a, if we have the high latency, it's like a download. We cannot, uh, sorry. It's a, like a outage. We cannot afford it.
[00:09:42] Andrew Zigler: Right.
[00:09:43] Sergei Liakhovetsky: From one side. So performance is one of them, definitely. And we are defining the core flows and we're looking on each and every core flow to see, upload, download metrics and also the single interaction [00:10:00] metrics.
[00:10:00] Sergei Liakhovetsky: Okay. So we are standardizing all of that. we're looking exactly every day, every minute, what's going on in the system on the other side. It's also soft metrics, like, uh, the sentiment of the customers. So we're running the service with our customers to get their sentiment, and when we see the sentiment is going up or down, we're analyzing it.
[00:10:24] Sergei Liakhovetsky: We're trying to figure out what's going on, and we're moving to the metrics and the system to correlate between the sentiment that we're getting to what we see in the system.
[00:10:34] Andrew Zigler: So that's very cool. So there's both trailing and leading indicators
[00:10:38] Sergei Liakhovetsky: Exactly,
[00:10:40] Andrew Zigler: this experience.
[00:10:40] Sergei Liakhovetsky: exactly. And it's always balance between both of them to see how we have leading, uh, leading the metrics like, uh, the sentiment of the customers, and really metrics to understand what is the happening in system.
[00:10:59] Andrew Zigler: [00:11:00] So we've talked a bit about how you create this environment of trust and how you make the customer experience first class from the very beginning of monday.com to still now its its scale and size as enterprise facing business. And now that we've kind of talked about maybe some of the less glamorous work of working on the foundation, there is a a, a pivotal time within monday.com that you. To me that was really fascinating that I wanna talk about before we move into, um, the different kinds of AI offerings that y'all have built out of that, um, experience. And this is your 30 day AI month, and it can be pretty hard to convince an entire organization to pause a roadmap for 30 days, especially at the size of Monday. But you pointed all of your technologists at one goal, you know, becoming an ai enabled org. And in this month, um, you experimented with a lot of things with your team that I would really be curious to know more about, like, one of them being like, how do you design a month like that? So people don't just build demo after demo after demo, but they're walking away with, you know, durable skills that are gonna fundamentally change how they [00:12:00] work.
[00:12:00] Sergei Liakhovetsky: so thank you for your question because it's a really interesting, and I think it will be. Create, uh, exercise also for other companies and other leaders in other companies. ' cause you know, before we, uh, moved or even thought about the AI month, it was like we worked with ai. We looked at it like, uh, features.
[00:12:22] Sergei Liakhovetsky: But it was like a table stakes. You have features, some teams are developing it, but it's not something that is transformational for anyone. Not internally and not externally. And with the leadership, uh, myself, who's another p and d, and also products, started thinking about it, what we can do differently.
[00:12:47] Sergei Liakhovetsky: How we can engage the entire builders company. It's like a company. We have about 700 people in builders, so 700 engineers. We need to engage, we need to inspire [00:13:00] and we need to give them some tools. For experimenting, hands-on, experimenting. Otherwise, if it's only education and training, it, uh, it's nothing for, uh, such a huge company like, uh, like we have in Monday here.
[00:13:16] Sergei Liakhovetsky: So we started thinking about it and, uh, we defined to ourselves several principles. So the first principle was we are not going to run it like a hackathon. Because when we're talking about hackathon, and hackathon is very important, vehicle mechanism to, to reach the inspiration, to innovate. But in our case, we wanted to achieve something else.
[00:13:46] Sergei Liakhovetsky: We wanted to make sure that whatever we're doing is reaching production. That we understand not only what to do, but also how to do that. So the first principle was whatever we're doing is going to [00:14:00] production. We're not going to do something only for the sake of experimenting here. The second one was, what are we going to do with our commitments?
[00:14:11] Sergei Liakhovetsky: We have customer's commitments, we have soft commitments, we have hard commitments. What are we going to do about that? And we started working with the teams. And just to, to give you some sense of, uh, the period of time it was about two weeks work. Okay. From the point we decided that we wanted to have AI months to the point we started, we kicked it off.
[00:14:36] Sergei Liakhovetsky: It was only two weeks
[00:14:37] Andrew Zigler: Wow.
[00:14:38] Sergei Liakhovetsky: and, and I think it's super important, and this is one of the main takeaways for me at least, like a lessons learned, if we decide to do anything, you need to do it right now. You, even if it works. Some mistakes if, if it's a mistake will happen after it, [00:15:00] we'll change and, uh, we'll fine tune as we're going.
[00:15:04] Sergei Liakhovetsky: So the second point was to go through all the teams and we have dozens of teams in Monday, and to make sure that we know how to deal with each and every commitment that we have to our customers. And at some point we figured out that. Most of the work that we're going to do, we can do also with ai. It's not like, uh, we need to stop everything and, uh, to start, uh, doing some esoteric things on only for the sake of learning here.
[00:15:38] Andrew Zigler: Right.
[00:15:39] Sergei Liakhovetsky: So this was the next principle that we define to our teams and to ourselves, that whatever we're doing, we need to make sure that it, uh, accelerate in our product map, even if we need to reprioritize, but at some point it should be fitting the product roadmap that we're doing. And [00:16:00] again, at some points we saw that it doesn't fit and that's fine, but those were like exceptions.
[00:16:06] Sergei Liakhovetsky: This is how we started working. This is how we kicked it, this is how we designed it. And after it, it was a huge amount of work to work with the teams and to run demos, like you said, and to have education. And we had the champions program, uh, in parallel to that's to make sure that people have
[00:16:28] Sergei Liakhovetsky: communication channels and know how to get the data and to get the knowledge they're looking for. It was from one side, it was a, a huge amount of work on the other side. It was so inspiring that people started, uh, actually fighting for going and to showing the presentations and showing the demos of whatever they're building.
[00:16:53] Sergei Liakhovetsky: And this was great. This is how we built. The new products, like, [00:17:00] uh, you mentioned before Monday Magic. This is where they started, Monday Vibe and also sidekick the Copilot that we're using the system and also internal projects like, uh, morphic for splitting them Mono Lead and Sherlock, where we succeeded to reduce the amount of, uh, time we are spending on tickets resolution by half, more or less.
[00:17:24] Sergei Liakhovetsky: So. If you give to people ability to fly, they're flying just to give them ownership and give them to run. I think this is the main takeaway from, uh, from the AI months.
[00:17:39] Andrew Zigler: I think that's a great lesson to learn from it. And I think that's a really hits at the heart of why developers are developers. Like we're software engineers because we want to build cool stuff because every day we want to go to work and build something interesting that changes lives. That you know, is, is is intriguing, but also just makes the world better. Like we're curious and we want to tinker. So the idea of having the ai [00:18:00] AI month and aligning it around some ground rules. I want to, I wanna run through them 'cause they're really smart. Uh, one of them being that anything that you put together, it's not just throw away hackathon.
[00:18:08] Andrew Zigler: Everything has to be aligned towards a business purpose, a goal, what was gonna be worked on, should have the intention of being taken to production, but then taking it one step further and working with all of the teams on an individual basis and understanding their needs and commitments to customers. Then you actually get an ability to map that into your experimentations with AI and what you're going to build and, and roll with. So now you're not just throwing like AI at the org and saying, figure it out. Let's, let's, let's learn how to roll with this. Let's see what we can make instead. Um, you're creating like a, an art of the possible.
[00:18:41] Andrew Zigler: You're, you're showing a space where people can come together and ship and show. Best versions of what they think the product could be. And I think that's the most exciting version of like AI month that can happen inside of any company is you get somebody who can take the the core idea, the market position of what you provide to your [00:19:00] customers and take it to that next level with AI as like a concept. It helps align everybody right around the idea 'cause right now we don't know what we're looking for yet. So I, I like how you, um, kind of used the carrot on the stick, so to speak. Like the guide, the experimentation towards what, um, the customers and stuff would ultimately benefit from.
[00:19:21] Sergei Liakhovetsky: Yeah, I think that, you know, people need some space to experiment. They, they, when they are looking for the mind shift, the mental mind shift and what we're going through with AI right now requires absolute mind shift in the way you are working, in the way you're looking at the staff, in the way you are practicing and in the way you experimenting.
[00:19:47] Sergei Liakhovetsky: If you'll not be able to experiment, if you'll not give teams to experiment. It'll be a failure. It'll be like, uh, like everything. And once you provide this ability, those tools to people [00:20:00] to fly, they're doing miracles. It was amazing to see all the speakers, all the demos, you know, overall we had allowed, around the 17 workshops only during this, uh, a nine months.
[00:20:15] Sergei Liakhovetsky: 17 workshops, we had about 22 speakers and we had, uh, yeah. And we had about 70 demos or something like that.
[00:20:23] Andrew Zigler: So just a huge number of participation. Everyone's really excited about figuring out what we can do with this and where we can take it.
[00:20:29] Sergei Liakhovetsky: Yes. And people really loved it.
[00:20:32] Andrew Zigler: this world where they're like using all these different tools, you know, are you just kind of letting them experiment and grab whatever they want? You can use this tool, you can use that. Like how do you start to keep tools and check and understand like, what is our tool library gonna be emerging from this month?
[00:20:46] Sergei Liakhovetsky: Yeah. So, uh, first of all, I think, uh, again, it's great question here because, uh, we started working with one tool, uh, actually with Scarcer, and we figured out that. [00:21:00] Limiting people is limiting their imagination and their ability to move fast and to experiment. So we defined absolutely different, uh, methodology and different processes here in the way we're working with tools, in the way we're acquiring tools, in the way we're experimenting.
[00:21:20] Sergei Liakhovetsky: So we absolutely remove the barriers. Actually, personally myself worked with security, procurement, and legal to define exactly how we're working from the one side to make sure that from security perspective. We're not, uh, putting our customers in threat. So we worked with data, for example, on the other side, we had zero bureaucracy policy here and zero bureaucracy policy actually, to make sure that in the one week you are getting all the tools you want to get, but you need to be champion of the tool if you are asking for this tool.
[00:21:59] Sergei Liakhovetsky: You [00:22:00] need to make sure that there is no PI thread there, and if after two months the tool is not used, we're just removing it for our catalog
[00:22:10] Sergei Liakhovetsky: and
[00:22:11] Andrew Zigler: you just basically kind of just like open the door, let anyone.
[00:22:14] Sergei Liakhovetsky: For anyone. Yes.
[00:22:15] Andrew Zigler: forward that they need to get their best job done and then, and trade away is, is that I love, I loved how you, I love what, how you smiled when you said, you know, it's a no bureaucracy organization. You have to own you, you own that tool, which is, I think is easier. Uh, it's pretty, pretty easy to do. If it's a tool you're really passionate about, you want to use, like, oh, I don't wanna be stuck using Cursor. I wanna use this other tool. It's like, it's, you're more likely to get like a, a good champion who can help others than to learn how to use and get the most
[00:22:41] Sergei Liakhovetsky: Exact.
[00:22:42] Andrew Zigler: Right.
[00:22:42] Sergei Liakhovetsky: Exactly. Exactly. And, uh, it, uh, made amazing effect. Amazing effect because people felt that they are owners and they wanted to make sure that the others are using those tools. So it was a lot of communication around it. It was a lot of talks, it was a lot of smiles there,[00:23:00]
[00:23:00] Andrew Zigler: Oh, I love that. 'cause then it's like, oh, this is my tool. I want to use this tool. So if I wanna keep this tool, I should convince
[00:23:05] Sergei Liakhovetsky: eh? Yeah. Look how it's amazing.
[00:23:08] Andrew Zigler: Yeah, exactly. Look at my demo, look at what I built. You know, I shipped this with this tool. This is the future. So, uh, it, what I love about this month you're describing is it's the right blend of incentives.
[00:23:19] Andrew Zigler: You have, the experimentation you have the, the career growth you have the alignment towards where the company's gonna go next. You're tapping into all of that excitement. But what I love that you just mentioned too, is that you also worked with security to keep things safe, that make sure there are boundaries.
[00:23:32] Andrew Zigler: You know, what, what did that look like? Just from like a bird's eye view of just making sure, um, you had the right fences up for that month.
[00:23:40] Sergei Liakhovetsky: So first of all, we worked with the app security team. Okay. So app application security team was part of the committee where we decided, uh, what we can do, what we cannot do, how to shape and design the MCP, for example, when we're working with the tools they were really engaged. Like, you know, not, [00:24:00] uh, every app security team is really engaged.
[00:24:03] Sergei Liakhovetsky: In this case. It was amazing, amazing effect. They worked side by side with development, like real builders.
[00:24:10] Andrew Zigler: So for someone that's listening to this and they wanna replicate an AI month within their own org, maybe they're in a similar like leadership position to yourself, what would you absolutely repeat and what would you change if you did it again?
[00:24:25] Sergei Liakhovetsky: So what I'll repeat is definitely, uh, look for results. Okay. It's not like, uh, like I say, the training or education for the sake of education. Whatever we define, like, uh, we're doing in these months, we need to define also what kind of results we want to achieve. The second one, I think it's, uh, decision to action.
[00:24:49] Sergei Liakhovetsky: It took about two weeks for us. And it sounds really quick, I would say from the make, from the decision making to the kickoff [00:25:00] of the months today, if I'll do it, I will do it in one, uh, in one week. Because the excitement should be like a boosted, okay. It's not like, uh, something that you need to work on it and to have.
[00:25:13] Sergei Liakhovetsky: And building, uh, all the processes around it. If you'll start building the processes when you want to boost the mental mind shift, it'll fail. So whoever is going to do like we did AI months should uh, make sure that from the point they decide that they all in on for it. It should be immediately.
[00:25:35] Sergei Liakhovetsky: The bottom up ownership. I think the, one of the most important things here, like I said, if you, uh, give people the ability to fly with the tools, with the decisions, with, uh, the methods, how they're working and what they want to achieve, they will do magic. They'll do miracles. So in this case, our work was [00:26:00] really easy as leaders, as the leadership team.
[00:26:02] Sergei Liakhovetsky: Just to provide people the zero bureaucracy, like we said, to remove the barriers and to give them to run forward fast. I think that one of the points that, uh, we need to make better next time is, uh, first of all, uh, the continuity. Whatever we start, we need to make sure that either we know how we finish it during this months or how we're dealing with that after the months is finishing.
[00:26:35] Sergei Liakhovetsky: So we had several golden initiatives where we continued working even after the months and uh, that's fine. Uh, but it moved our priorities, it moved our scope. And they took some time. So once you decide what you are going to do, you need to know how you're going also to finish it. And again, it, uh, that, that's fine to do [00:27:00] mistakes, especially when you're doing something first time.
[00:27:04] Sergei Liakhovetsky: I think that next time we'll think about the scope, uh, a bit better there. Okay, so this is more or less the the points that I wanted to to mention here.
[00:27:16] Andrew Zigler: No, it's really useful. I think a lot of folks would be able to take these into the, their own future experimentations. I love the idea of like, give constraints, um, but then also align it towards like we're gonna have closure. Right. We're gonna either ship
[00:27:29] Sergei Liakhovetsky: Yes.
[00:27:29] Andrew Zigler: know what happened with it, and we're gonna close the book.
[00:27:31] Andrew Zigler: But don't just experiment without the constraints. Don't experiment without coming back to see what those experiments did. I think that's something that we talk about a lot here on Dev Interrupted, um, about like measuring and understanding the adoption and also the impact, um, of those tools and initiatives.
[00:27:46] Andrew Zigler: And so I wanna talk, um, just in our last segment here, about like the result of that AI month. And it created this whole ecosystem of products. And now I understand why you've taken us on this journey. Everything was constrained around this needs to align towards a [00:28:00] customer usage. This is gonna be something we take to, uh, this is something we're gonna ship as a product.
[00:28:04] Andrew Zigler: So now it makes sense that there's four offerings and you have magic and vibe and sidekick and Agent factory. And maybe on the surface, maybe it sounds like I know what they do, but assume I don't. You know, what makes these four, uh, tools like fundamentally different from each other?
[00:28:21] Sergei Liakhovetsky: So, uh, actually those are different tools and for different, uh, purposes and intents of the user. So when we're talking about Monday Magic Monday Magic is how to build solutions when you build the first solution with Monday. So it's, uh, for someone who is the builder, starts to build, for example, some solution like a library in the university or, uh, to, uh, create a shifts in a hospital or anything else.
[00:28:53] Sergei Liakhovetsky: Okay. Or solutions with CRM, for example. So this is about man magic that you are working with the prompt. [00:29:00] And you are getting reference implementation here. The second solution Monday Vibe, is where basically you can, uh, build the applications, uh, that, uh, are relying on uh, Monday entities on Monday boards, on Monday dashboards.
[00:29:19] Sergei Liakhovetsky: And uh, I think this is huge because whatever you are doing, you don't really need to know Monday we are just work in the Monday environment. You're building the applications and you can continue working with various applications later on, and we see a lot of traction around it. Uh, the third one is a sidekick.
[00:29:39] Sergei Liakhovetsky: Once you already build solution and build applications, you have a copilot, it's a horizontal copilot, so use a user or, or is a user or is a builder of Monday, you can use sidekick that will continue helping you to do whatever you want to do in the system [00:30:00] to fill the boards with the items, to remove the items from the boards to connect between different boards.
[00:30:07] Sergei Liakhovetsky: You can ask through prompt Sidekick to do that and sidekick will do that for you. And the, the last one is the agent factory. Maybe not the last one. Maybe I missed several. The, we talked about AI columns and blocks. We doing a lot of work. DI here, but we're not talking about. Agents factory. So basically, agent Factory allows you to build the vertical solutions and when you're doing, uh, building vertical agents, those vertical agents can work with Sidekick.
[00:30:38] Sergei Liakhovetsky: So you'll have exactly the same context. You'll share the context of Monday, between Monday count, between different AI solutions and AI tools and when you're sharing the context, actually you have like a compound effect here that is bringing a lot of value to our customers. So I [00:31:00] think this is exactly what we're looking for, and we'll continue in region, our portfolio of AI with more products and more solutions.
[00:31:09] Sergei Liakhovetsky: So altogether will provide us the absolute compound effect and help to our customers.
[00:31:16] Andrew Zigler: It is really fascinating to listen to you describe it because it's like these different levels of AI and your comfort and technical level familiarity and you know, most companies are just kinda shipping one thing, but you saw in your user base that a one size fits all solution wasn't going to work.
[00:31:33] Andrew Zigler: And from how you ran through it, like magic is, it's like prompts. It's. It's, it's simple prompts to get simpler things done, and then you can go a level higher, right? With Vibe, you can kind of orchestrate these apps on top of the platform, on top of the tool, and then you can kind of work with Sidekick.
[00:31:48] Andrew Zigler: Now you have an assistant, right? It can, uh, understand everything within your monday.com world, um, and probably trigger a lot of these other workflows from it as well. And then you have Agent Factory, which is like a level of abstraction above that. [00:32:00] So no matter how technical you are, uh, or aren't, you can come in and pick up one of these tools.
[00:32:06] Andrew Zigler: And make it fit for you. And it's gonna make Monday fit, like a glove fit, like a glove for whatever you need it to do, or however you log into monday.com. Right.
[00:32:14] Sergei Liakhovetsky: Y Yeah.
[00:32:15] Andrew Zigler: Really fascinating. Um, is that kind of
[00:32:18] Sergei Liakhovetsky: It,
[00:32:18] Andrew Zigler: you saw in your user base that led to creating that suite?
[00:32:22] Sergei Liakhovetsky: yes. And the, uh, this is exactly the point, uh, because, uh, you know, we are learning our users and the Monday platform is very widely used in different, uh, industries and for different markets. So. Okay. Once we, uh, providing vertical agents on one side, the solution on another side and horizontal sidekick, co co-pilot on the, uh, the short vector, basically we provide our users with the ability to decide how to work on one side.
[00:32:56] Sergei Liakhovetsky: On the other side, we are learning about the user to [00:33:00] improve the context and improve, uh, the experience because we understand the intent of the user. And, uh, knowing the intent of the user and knowing the data about the user, uh, brings a lot of value for the user itself because we can, uh, provide more better quality for, for, uh, the agents that we're building for them.
[00:33:23] Andrew Zigler: No, I, I love that bit. You called out about how it helps you understand what users come to you for. Why are they using Monday? It's like just as much as they get more value out of your platform now you also now get to more intimately understand why they use you and why you're sticky for them and how you can meet them where they're at and provide things that they don't even know that they need yet. And uh, I think that's really fascinating. I love how you had mentioned how they kind of compound on each other. That's something that stands out to me is if you have these different ways between them, you can triangulate some level of technical familiarity and domain level expertise to execute something with these tools.
[00:33:59] Andrew Zigler: [00:34:00] So how do you create the boundaries when you're engineering in that space? I imagine in the beginning maybe it was a little fuzzy, but as they've come to the come to be like full featured products, what's this mental model within monday.com that separates like, what is a sidekick problem from this is a vibe problem.
[00:34:17] Sergei Liakhovetsky: Vibe problem is, uh, to give use user ability to build the application that you're looking for. Okay? You already user in the system, your building, the applications. You are using those applications. You can publish those applications for. Other users to use in the account. Okay. When we're talking about sidekick, it's a different, uh, different dimension.
[00:34:44] Sergei Liakhovetsky: You already working with your entities, you are working with your dashboards, with boards. You already gather the information that you want when you build this application. And now with Sidekick, it's like a horizontal [00:35:00] copilot in, uh, other places, in other applications. You can do whatever you want in the platform level.
[00:35:08] Sergei Liakhovetsky: Okay? But in the platform level, you need more intent. You need to understand more context and you are getting more context from, uh, the vertical solutions like, uh, CRM agent, for example, work management agent. It's a vertical solution. Then understand the customers, and this is how you define the boundaries.
[00:35:32] Sergei Liakhovetsky: So sidekick, more horizontal. One a vertical agents were building with agent factory. Okay, and the applications we're building with Vibe that can work in the scope of board or in the scope of applications or in the scope of account.
[00:35:48] Andrew Zigler: So what, what does this do to your, like infrastructure that we talked about at the beginning, right? You did all of this hard work of making sure the foundation underneath this was sturdy and steady with all of these, um, [00:36:00] agents and all of these AI features on top of, of your code base. Um, like. Did you encounter more constraints or problems with like your API? Like suddenly, oh, we're getting way more requests. And maybe it was fine because you spent all that time at the beginning preparing for it. But like, what did, what did you see change internally in your infrastructure once you started to roll these things out?
[00:36:20] Sergei Liakhovetsky: Yeah, so definitely agents are, uh, interacting with, uh, with the system absolutely different from, uh, human beings. And we already see it. We already see that we have fan out of engines. We have a lot of API calls that, uh, we're looking and see how we're going to have a fairness index, for example, to make sure that the one account is not, not taking all the resources.
[00:36:49] Sergei Liakhovetsky: We're looking also on concurrency. To make sure that the agents get the fair resources. And again, we're moving when we're working with, say, with ai, we're [00:37:00] moving from SaaS that is CPU bound to SaaS that is GPU Bound is absolutely different. It's like, you know, like from one side are getting resources like eh.
[00:37:13] Sergei Liakhovetsky: Expensive, like, uh, Ferrari. Okay. On the other side, uh, you basically need to know how to manage them. So we're building the, uh, guardrails. We're building the schedulers to make sure that we know how to manage the costs here. So it's not only about how to provide the resources, but also about how to manage the cost
[00:37:37] Sergei Liakhovetsky: of first resources and they, I think it's a absolutely different way, like, uh, to, to look on the metrics and to look on the SLOs. It's different type of SLOs here. When you are looking on the cost on fairness index, when you are looking on the concurrency and the agents utilization and how to deal with fan out here.
[00:37:59] Andrew Zigler: [00:38:00] So was this like, it opened up a door? It's like a, like a, like a new world of reliability that you had to offer because now suddenly. Your engineers are building differently on your platform. Your users are using your platform differently, and now there's this new cohort of users, these AI agents and bots that are also slamming your APIs and systems.
[00:38:18] Andrew Zigler: Right? So, um, really it sounds like from all of that, you couldn't just innovate and go really, really fast with all these new products, but you also had to, you know, going back to the very beginning, make sure you had the right foundations that to, to take them at scale for how monday.com customers expect.
[00:38:35] Sergei Liakhovetsky: Right foundation and also right security guardrails. Uh, we didn't talk about it before, but, uh, in a new world of ai, we need to make sure that, uh, our customers are protected, that we have, uh, right segmentation of the network, that we have a right guardrails around the count isolation here. So it's a absolutely different story [00:39:00] here.
[00:39:00] Sergei Liakhovetsky: Whatever we had so far is changing right now, and we are changing together with that and the infrastructure is changing. Working with GPUs and working with the new way of work and new way of interacting with the system like agents are doing, we need to build different APIs. We need to optimize those APIs for MCP use and for agents use.
[00:39:26] Sergei Liakhovetsky: It's a different story at all.
[00:39:29] Andrew Zigler: Yeah, you're just like talking about this new frontier of all these things that you're gonna build. And honestly, in talking with you, Sergei, I get the vibe that like, we're, you're gonna, we're gonna end this call and you're gonna go back to building it. It's like you're such like a builders person. and I can tell you're so passionate about bringing this stuff to production and it's a really amazing.
[00:39:44] Andrew Zigler: To hear this story, you know, you've taken us inside of Monday's Foundations and their AI ecosystem journey, how y'all evolved, your 700 technologists organization to become AI ready, but also aligned around goals, uh, making products that your customers would actually use. And I think that [00:40:00] I learned a ton of stuff about how I would model this internally.
[00:40:02] Andrew Zigler: I know our listeners did as well. But before we wrap up, where can our audience go to learn more about you and the tools from Monday?
[00:40:11] Sergei Liakhovetsky: So first of all, from a monday.com of course. Uh, we have a great blog. We have great articles and, uh, everyone is welcome to visit our site to learn about our tools. We have a lot of, uh, articles and blogs that we're publishing also on LinkedIn. As a pause and, uh, I will be happy to answer any question of, uh, of the users of the, of this episode.
[00:40:38] Andrew Zigler: Awesome. Well, we'll include those notes in, um, the show notes. That way people can go check out those links. And thank you to everybody listening today, but the conversation, it doesn't end here like Sergei said, we want to continue it on LinkedIn Please come find us if you have questions, uh, curiosities concerns, anything about what we talked about today.
[00:40:55] Andrew Zigler: We would love to know about how you are using AI within your org, but also what you're taking away from this [00:41:00] conversation. And so join us on LinkedIn. You can also find us on Substack. Just look for the Dev Interrupted newsletter. And that's it for this week. See you next time. And Sergei, thanks again for coming on the show.
[00:41:11] Sergei Liakhovetsky: Thank you for having me here. Thank you, Andrew.
