Data to Decisions: Using Observability and Automation to Boost Business Outcomes

[00:00:00] Chapter 1: Introduction to Code RED Podcast

Mirko Novakovic: Hello everybody. My name is Mirko Novakovic. I am co-founder and CEO of Dash0. And welcome to Code RED code because we are talking about code and Red stands for requests, errors and Duration the Core Metrics of Observability. On this podcast, you will hear from leaders around our industry about what they are building, what's next in observability, and what you can do today to avoid your next outage. So today my guest is Bhaskar Sunkara. Bhaskar is the co-founder and CEO of Bicycle, a 24-7 AI tool that analyzes business operations data and automates actions based on that. He was also one of the co-founders and founding CTO of AppDynamics, which Cisco acquired in 2017. His core background is observability, and I'm really excited to have him here on today. Bhaskar. Welcome to Code RED.

Bhaskar Sunkara: Thanks for. Thanks for having me.

[00:00:57] Chapter 2: Recollecting Past Experiences with AppDynamics

Mirko Novakovic: We know each other from AppDynamics time very early. Yeah, because I was one of your first resellers or the first reseller in Europe. And we basically had a lot of trips around Europe to customers. And with regards to my first question, what was your Code RED moment? We could share a few from our customers back then.

Bhaskar Sunkara: But a lot of war stories, a lot of war stories. Yeah, absolutely.

Mirko Novakovic: What's your best war story?

Bhaskar Sunkara: I think this was actually during the AppDynamics journey. We had built the tool. We were kind of in the early days of selling it, and this was basically like a gaming or a betting company. And the situation was that they were setting us up on their new application that they were building. But the current application, which was the core business you know that was basically really running day to day and doing business that didn't have, you know, AppDynamics on it. And they just had an outage on it. And someone basically made a change and just blew things up. And then that basically led to them losing a massive amount of business. You know, so it's down for a day or half a day or something. But it was, you know, they lost almost, almost $1 million. So for me, it was a moment where it really quantified, you know, what does observability, what does monitoring actually mean? You know, in terms of, you know, because you because you kind of know what's really happening. But once you really have real time scenario once you have a real scenario, then you really know what it actually means.

[00:02:32] Chapter 3: Career Beginnings and Observability Innovations

Mirko Novakovic: Yeah, absolutely. And starting in your career. Back then, I watched LinkedIn and I saw Pramati, which I remember very well because I was working on WebSphere and pragmatic was basically a competitor out of India. And I remember that it's a Java application server. Tell me a little bit about it. How did you get there?

Bhaskar Sunkara: Absolutely. It was definitely one of the that was my first gig. It's where I started off. It's where I really got to sort of pick up the skills that were handy for the rest of my career, right? So I started off with middleware, right? Like so. And this was the time to your point, they were about 20 to 25 application servers. It was kind of the next big thing before it got commoditized. So we were started out of India, and then we had like a really amazing team, and we were just so focused on creating technology because Sun did a pretty good job of, you know, writing out the specs of what is an application server, what are the different services in it? You know, here's sort of how you do resource pooling, here's how you manage transactions and all of that. And it was just fun to, like, really translate all the specifications into a piece of software. And then they had a compatibility test suite and that's how you test it. So yeah. So first job was a lot of fun. I spent about three years just sort of focused on coding. And then the two years after that, I spent about five years there before joining Wiley. I actually spent it on the field. I started working with customers doing a lot of pre-sales, but I also had my coding job, so it was a lot of fun looking at sort of things going on in the market, understanding what it takes to deploy a product. How does it sort of, like, really work with respect to doing a deal? But also had sort of the say in engineering on what should we put into the product, you know, based on what we're hearing.

Mirko Novakovic: Yeah, absolutely. And it was also a core competency, I think, at AppDynamics, which I experienced, that you were you and also Joe, you very customer focused. Right. And talking to customers a lot. But then you joined Wiley, basically the inventor of APM for modern. Absolutely.

Bhaskar Sunkara: Absolutely.

Mirko Novakovic: Yeah.

Bhaskar Sunkara: Yeah. So Lou Cerny was the founder and he actually has the patents for bytecode instrumentation. So that's, you know, that's going back a long way, but yeah, it was awesome to join at that point.

Mirko Novakovic: Yeah. You worked on the BCI part. Right. On the bytecode instrumentation part. Yeah.

Bhaskar Sunkara: Yeah, I was on the team building the agent, so it's almost like the transition went from building middleware, which is all the plumbing required to, you know, actually execute transactions and resource pooling and all that stuff to, to really monitoring it, you know, and and it was, it was kind of an advantage because once you understand how transactions flow through in an application server when you get to transactions and when you get to monitor them, you, you, you have that knowledge. So I think it was an advantage for me as I got into the actual sort of BCI agent team, because having knowledge of that layer really helps.

[00:05:44] Chapter 4: The Dawn of AppDynamics*

Mirko Novakovic: Yeah. And that's also where you met Joty and and you guys co-founded AppDynamics back then and one when I, when, when I met you the first time. Pete actually a friend we both have, and my co-founder at Instana, he was one of the first non-technical guys on the or the first non-technical guy on the dynamics team. And he reached out to me when you were 6 or 7 people and I came over to San Francisco. And what really impressed me, coming back to bytecode instrumentation, that AppDynamics was the first tool where you did not have to manually put in the right information. So going back to that time, the way tools like Dynatrace or Wiley or others work were, that you basically had to select which packages and classes and methods you wanted to instrument, right? But with AppDynamics that worked magically out of the box. And I think that was like, I was so impressed. Can you.

Bhaskar Sunkara: Yeah.

Mirko Novakovic: Tell us a little bit about that, that invention or how you came up with it?

Bhaskar Sunkara: I mean, you know, you and I actually talked about it a lot of times. And then when, when you guys started reselling, that was one of the things that you were the most curious about, because, I mean, you had seen the market? Or how much pain it is to basically say, oh, you know, like these are the packages that are important for me. I changed some code, I added a new package, so I have to put that into the config. Not only that, you have to restart everything because the JVM would pick it up on the restart and instrument everything on startup again. So there were, there were a, there were a few new APIs that Java 1.5 and 1.6 basically introduced. And they had to do with one being able to determine how, you know, when a transaction is happening, what are the classes involved in it? You know, so that was one of the APIs that we did a lot of research on. The second API was basically being able to retransform those classes so that you don't have to restart the JVM. So those two changes really gave us the edge you know, in the market when we went out there and really disrupted the market based on how much config you basically had to do, and we could all do it out of the box. We had the aspect of discovering what are the packages, what are the classes involved? What are the meaningful ones? Which are taking a certain amount of latency and we're able to instrument them and without restarting the JVM. So that was a game changer in terms of you know, observability and configuration and, or, or rather taking away configuration.

[00:08:23] Chapter 5: AppDynamics' Rapid Growth and Acquisition by Cisco

Mirko Novakovic: Yeah, absolutely. I mean it was really a game changer. And it also, I mean, it led to a really impressive traction of the company. Right? I remember the first years you blew it away, right? Growing super fast, super aggressive. Yeah. With some also amazing sales people on your team and over time, also getting new, really great sales leaders who are now crows at really large companies.

Bhaskar Sunkara: Yeah. Yeah, I mean absolutely. I mean.

Mirko Novakovic: Yeah. And then you sold in 2017 AppDynamics to Cisco for $3.7 billion. So one of the biggest software acquisitions at the time, right at that time.

Bhaskar Sunkara: It was huge at that time. It was really huge. Definitely. Those numbers were kind of rare.

Mirko Novakovic: Yeah. And also I remember because you were I don't know the story exactly, but I think you were a week before IPO when that happened. We we were all prepared that you go IPO and celebrate at Nasdaq and then bam Cisco took you. Right.

Bhaskar Sunkara: Absolutely. I mean, actually the team was in New York to ring the bell. And so it was pretty dramatic. And this was the time when the elections were done and the change in government and all that stuff was happening. It was pretty dramatic in how it happened at the last minute. Everybody had bought suits and they were basically there. And then Nasdaq felt so bad that they basically got us to come in and ring the bell after the acquisition and Sysco and us, we went in and did it and it was a lot of fun, but it was pretty dramatic at that time.

Mirko Novakovic: Yeah, absolutely. And then you stayed at Sysco for a while, for a few years, and I watched it. So you had a lot of interesting AI features added to the AppDynamics platform. It was growing, and I think that's also a little bit of what you're doing today. But I saw that you shifted a little bit more to business. So there were these features like business IQ and other things. So I think you learned that observability is not only about technology, but it's also about the business, right?

[00:10:32] Chapter 6: Transitioning to Business-Focused Observatory with Bicycle

Bhaskar Sunkara: No. That's right. I think I would say that business IQ, which was one of the products that we did on top of the core APM strategy that we had for, for AppDynamics that, in fact, was probably the in between the stepping stone to kind of what you know, we're doing at Bicycle right now. What we kind of really thought about was and this was from multiple conversations with customers. One of the conversations one of our really old time customers who used to give us a lot of feedback they basically said that. Look, I mean, you guys are, like, really looking at outages and things going wrong, etc., but can you also do the impact of an outage? So for example if an outage happened, can you get the names of everybody whose transactions were impacted and give me a report so that I can actually send them, you know, like basically cards or gift cards or something for the issues that they face. Right. And that's when you actually start thinking about you can go from whatever has happened to basically the impact of it and understand the impact of it at a business level. And it's just not a way of just safeguarding technical KPIs, but also going to the next level where you actually are able to look at, you know, the business impact. So that was the business IQ inspiration. And then we built out a product that's kind of based on those principles.

Mirko Novakovic: Yeah. And let me let me explain. So a few weeks ago you gave me a demo and I have to say that I was really blown away. Right. And honestly, because on the one hand, I'm a little bit skeptical about all the AI stuff happening.

Bhaskar Sunkara: Yeah, yeah, yeah.

Mirko Novakovic: But then you showed me, and I will just explain with my words what happened, right? You told me, please give.

Bhaskar Sunkara: Me. Yeah.

Mirko Novakovic: Give me a website name of a business you want to monitor. Right. And I told you Edmonds.com, which is a company that sells cars like e-commerce. Right. And based in LA. And I told you the exact.

Bhaskar Sunkara: Customer x Instana customer. Right. Like so. Exactly. Exactly, exactly. Yeah, yeah.

Mirko Novakovic: Yeah, exactly. So I had knowledge. So I said take Edmonds. So you put in Edmunds.com and then what you do is it basically scrapes the website or does something looks at the website and then after a few seconds, it came up with suggestions of use cases and KPIs for that website. Right? Really business KPIs like number of cars sold average sales price and and really, really into depth. And then basically from that metrics you provide sample alerts and thresholds and KPIs that you should monitor. And then you also suggest how to map them to the systems where you can basically get those data from. Right.

Bhaskar Sunkara: Yeah, absolutely. So yeah, the way I describe it is that like Bicycle focuses on really the automation of business operations. Right. So at the core, it's an analyst co-pilot. So you give it operational data and it will look at what is the fastest path to convert that data into insight into action. So, so it's kind of laid down that path right now because it's not about viewing data on a dashboard. It's actually about using that data actively when your business is running. And the, the, you know, if you look at like the, you know, AppDynamics it's really focused on technical operations. And if you look at like what Bicycle is doing, it's focused on business operations. And the key thing is that the KPI that you're anchoring on and the KPI that you're really focused on is very different. Right? Like you focus on technical KPIs. It's sort of, you know, like the acronym you have for Red, right? Like it's basically, you know, errors, you know, duration, latency, all that type of stuff. They're very horizontal type of KPIs. But when you do business operations, it's a top down strategy you're focused on to your point number of cars sold, average price searches for cars and things like that. Those are the KPIs that guide business operations. Right. And then once you have those you know, you're really figuring out like, what is the dimension for that KPI? So our search is off in a particular city, you know, our number of cars sold in a particular city or for a particular segment for a particular you know, area.

[00:14:56] Chapter 7: Bicycle’s Innovative Approach to Business Metrics

Bhaskar Sunkara: Is that an issue or is it sort of going up? It could be a leak in the revenue. It could be an opportunity in the revenue. So that's at the core of what business operations is, right. So if you look at how you can automate, you know, business operations. So you have data. And that data basically can be multiple things. It could be click streams of people searching for cars. It could be orders going into a warehouse. It could be some issues that they're facing where things are going at a log saying, hey, I couldn't sort of like search for this and something happened because you know, of this particular error. Now you have all this data. But if you want to start looking at business operations, what are the automations that you can basically do, right? Like number one, based on the KPI and the context that you need to really like work with? Number one is like, what is the data you onboard. Right. So you have to get the right type of data. You pick the right event, you pick the right dimensions for that dimension.

Bhaskar Sunkara: And for that event, you have to pick the right logs. So it's just all like really putting all that stuff together. So that's step number one. How do you onboard you know data. Number two is once you've onboard a data now you're ready to actually look for patterns in the data. Right. And you can automate, you know, a bunch of things in it. But what pattern do you look for? So that's the second piece to automate, right. And the third piece to automate is basically actually running the machine learning algorithm, which is basically going to look for not anomalies in the technical parts of it, but basically sales for cars in particular region. And each region has to have that baseline because you can't generalize it. It's a business baseline versus it's a technical baseline. So that's the third part. Then once you applied ML and found the pattern saying, hey, something's off with this region. You then have to automate the identification of the cause on why is this basically happening? Is there a technical reason? Is there a business reason? Is there an external reason which could be macro or could be something else related to that? And then finally you can automate action, right. So there's an opportunity to, you know, really automate a lot of these. And I think one of the things that was like a great insight for us was we're using the power of LLLMs to function as a business analyst for a business.

Bhaskar Sunkara: Right. So the business analyst has the top down strategy on, hey, you know, this is what we need to onboard because these are the KPIs that are the most important. It knows enough about the business. These are the patterns you should watch for. And then hands over to what we call a Bicycle data analyst. So, think about, like, the Bicycle business analyst. The Bicycle data analyst. One is rooted, one is grounded in AI and LLMs which is providing the context. And the other is basically grounded in machine learning and really understands how to crunch data, how to find patterns and things like that. And then they work together. And that's the whole Bicycle solution. But again, to your point as you were talking about the examples of what we did with the Edmunds.com use case, it's really the business analysts rooted in LLLMs specifically trained for verticals. So we actually focus on transactional businesses like retail and travel and fintech. And then it gives you the strategy of operationally, what are the KPIs to track. And that's how the whole system was set up. So hopefully that kind of made sense. I can drill into any of the pieces of it with abstract.

[00:18:29] Chapter 8: Automation of Business Operations with Bicycle

Mirko Novakovic: I have, I have really a ton of questions around that. So let's get back to that. Because I saw that exact metric. So you said number of category, for example, SUVs sold in Chicago, for example. Right. In a special location. How do you extract that? Is that something that you have pre-configured because you have knowledge about the car industry, or is that something you extracted from the website by using an LLM analysing the text on the website? How do you get there.

Bhaskar Sunkara: Along with data? One of the things we use is what we call a knowledge base, right? So if you, if you, look at like what LLLMs are capable of doing, they're able to do research on the web and compile a bunch of information. And so given a company, we can do that sort of any time. And that comes back with here's the key KPIs. It has to be very targeted. You know, it depends on really what you want to do. Like if you want to write a blog post or a company, then what you compile or what you look for, what do you do research for is very different than us focused on, you know, business operations. What are the KPIs that are important? What's the business model of the company? So there's kind of enough out there on the web. And the LLLMs have a lot of that data. It's really about like, how do you, you know, get to that right piece. But there's also like some sort of fine tuning based on verticals. So that's one of the reasons why I mentioned that we focus on you know, retail and travel and fintech because you can give the system the context on these are transactional businesses at the very core. The transactions can be bookings, payments, searches, etcetera, etcetera. So just that context keeps building up and the layers kind of build up to the point when you add the data, when you add whatever you've compiled for Edmunds.com, it makes it even more specific.

Mirko Novakovic: Yeah. Totally. Totally. And then because you now know that you want to track those numbers you've sold in Chicago, you have to connect to a system that provides you the data. Right? So you help with that too, right? You basically help mapping to, I don't know, a Salesforce which has the data. Right.

Bhaskar Sunkara: Yeah. Absolutely. So again, think about how that could happen. So let's just say Edmunds is more of a real time system. So some businesses care about real time, some don't. Right. It really depends on how frequent your transactions are. If your transactions are sort of like more on a weekly basis, then, you know, you can just do a weekly analysis. But if your transactions are really high, high frequency. So let's just say at once basically has high frequency transactions that are streaming into Kafka. Right. And one of the one of the topics in Kafka is basically streaming the, the searches and the orders that are coming in for Edmunds.com. Now, usually the streams don't really have the business metadata, right. They don't have, you know, the city, the value of the customer, the metadata of the customer. They don't really have business context. They'll have like a customer ID, they'll have like a car ID you wouldn't know whether it's an SUV, you wouldn't know what brand it is and things like that. But the business wants to work with the business context of it all the time, or the queries. The operationalizations need to be that layer.

Bhaskar Sunkara: So what we do is we can use the power of LLMs to really say, hey, we need to track the number of cars sold in Chicago. You know, just going with that example. And so what we'll end up doing is look for the event, which is the representing the sales. Right. And again the LLM has really good intelligence to pick up. Like what is the stream of all the streams in Kafka. So you, you ingest all that data, all the metadata and figure out what is the stream that basically is doing this, both by looking at the name as well as introspecting the data. Right. And then if you also, to your point, connected us with the CRM, we will then do a mapping of this customer ID or this car ID or this brand ID basically is mapping to this brand in the table. So we do that work ourselves versus basically asking you to say, can you map this? Can you map this, can you map this? That's what makes the next step which is onboarding of the data. And then we set up a normalized structure on which we can then set up a whole bunch of pattern detection.

Mirko Novakovic: Yeah, totally. So you basically map the concept of observability to a higher level, to the business level. And now what you call the co-pilot, the business analyst basically helps you automatically analyze your streams of business data and figure out what something is wrong and even give you hints where it's going wrong. Right. And why.

Bhaskar Sunkara: Yeah. So again, you know, if you go step by step, number one is like automating, like what are the KPIs that are important. And you know we give you recommendations and you can pick from it. Now second is onboarding the data. Three is set up the patterns. You know that the algo you want is sales by city right. And you set up like a threshold. You set up like an automatic baseline for all cities. You account for seasonality. You account for all of that type of stuff for the cities. And now you have an algo that's running all the time, whether it's running on a, you know, you know, minute by minute basis, whether it's running on a day by day, day to day Basis. And now you've suddenly realized that sales in Chicago are down. Right. So what happens next? Right. Like, that's the key thing. And usually that's where dashboarding tools basically stop. They're like, hey, something's interesting. Also, you have to be looking at the dashboard at the right time to be doing that because again, there's no automation. Once something is off, you know, why is it really happening? And for us, when we do the cause analysis, we break it down into three different categories. You know, so number one, it can happen because of a tech factor.

Bhaskar Sunkara: Maybe sales in Chicago are down because something's off with the, you know, application. Something's going on where people could not sort of you know, do software transactions. And there was some issue, right, with the service that's preventing it. So that's one. The second one could be a business issue. Maybe the way you're sourcing prices and the way you're setting up prices in Chicago is too high, right? And basically, people are like, you know, that's that. That's too high for me. And that's a business reason. It's got nothing to do with the technology working in the background or the reasons could be macro. Maybe you have a new competitor that introduced services in Chicago, and maybe that's what's really happening. Right? So for the first time now, especially again, back with the sort of like AI and how powerful it can get. It is not just about your corporate data, but it's also about what you generally know you know, about that particular factor, right? Like, so you could not bring that factor into a dashboard saying, hey, there's a new competitor for AdWords, and they're basically introducing services in Chicago. And that's why your sales in Chicago are down. But it's a perfectly acceptable, you know you know, cause and then you, you get to work on it, you know, like whether you do marketing, whether you do whatever.

Bhaskar Sunkara: So that's the automation of the cost identification. And what we do there is we have an agentic workflow that basically looks at what are the possible causes that could happen. And these are tech causes. These are external causes. These are business causes. Looks for each one of them analyzes data and tries to find potential issues that could happen in that. And that is eventually what leads to Actionability. If it were a tech cause, that was because of some changes that were done in a recent release. The action is roll back the release, right? If it were a business cause where there's some issue with the supply, there's some issue with how you're putting up prices and how you're calculating prices. You fix the business logic, you don't fix the technology then, because you know that's not your fault. If it's macro and you have to react to it as a business, then you do that. But that's where it creates a pathway going from onboarding of the data to finding the pattern to analyzing the cause and then taking action.

Mirko Novakovic: Yeah. It's interesting. And for the first category, for the technical problems, you probably could connect to observability platforms to get data.

Bhaskar Sunkara: So yeah, basically for tech causes. Right. We integrated with all the, you know major observability platforms, including including AppDynamics, of course, but but what we do there is that once we know it's technical this, you know, again, from data, there's a lot of things that you can already kind of pick up, which is what service is going wrong. Right. So, for example, in the data, if one of the dimensions is service, if one of the dimensions is like a particular shard of a database or something like that, we can look at the, the, the the events that are causing that pattern and triangulate what part of the technology and what that what what I mean by that is that you're able to say, hey, most of the issues that we're seeing is with this service, right? And so then you can actually create an action where you can drill down into that service and give them a deep link into that service dashboard for, you know, for Datadog or whatever you're kind of using. And so, so we, we integrate pretty deeply with observability tools because anything in the tech layer, you are kind of really going to that layer.

[00:28:05] Chapter 9: Technical Foundations and Challenges

Mirko Novakovic: It is OpenTelemetry a thing for you to use that as a integration layer.

Bhaskar Sunkara: Yeah, 100%, man. I feel like, you know, OpenTelemetry I think as it picks up more we in fact, like the original thesis of how we built data collection, we built it on the OpenTelemetry model where everything's sort of like a span and everything comes together into a transaction because even with data. Right. Like, it's fragmented. So what happens is sometimes and these are all real scenarios. Data from North America is in a different topic. Data from America from India is in a different topic. You know, the web requests of someone searching for something is in Google Analytics. The booking is in BigQuery, right? So this fragmentation all across and you have to kind of stitch it together. And that's the philosophy of OpenTelemetry, where you basically are getting a lot of these spans and they're kind of standard. But then you kind of put it all together to basically say, hey, here's the story for the transaction, here's the story for. And so I kind of apply the same principles to the data piece also. It's as fragmented as, you know the way you look at a distributed application.

Mirko Novakovic: Yeah, absolutely. And I like that you basically take OpenTelemetry to the next level, right? By providing you call it a story, which I like. Right. Basically giving it a story, a business story about where it belongs to. Right. In the context of the business and that's very, very valuable. And so I'm trying to imagine where, where this really fits into an enterprise company. I mean, you would normally if I wouldn't have Bicycle, I would. I would see that I would have to do this all manually. Right? I would have to put that data into something like snowflake or whatever and then build my own dashboard, do my own training. So you basically do this whole process of basically ETL, data warehouse analytics, machine learning, and in one tool, right?

Bhaskar Sunkara: Yeah, absolutely. I mean, think of us as like a force multiplier for data teams, right? Because all the work, all the things in the background that they need to do you know, a lot of it is homegrown tooling because viewing data, there's a lot of work done in data platforms, warehouses and things like that. Right? Like the core data infrastructure of how do you query better, how do you join data together better and things like that. But the consumption of data by the business is where tooling really relax, right? You know, viewing, you know, viewing of data is basically all of the dashboarding stuff. But how do you use that data? How do you consume that in a timely manner, and how do you automate the consumption of it? That's where tooling really. So that tooling really limits data teams as they, as they, as they work on it. So we just act as a force multiplier by giving them the solution. You can use it both for internal consumption where you're like, hey, I'm using it for operations saying, how do I do this part better? How do I make it more efficient? How do I optimize inventory, etc., etc. but you can also do it for external consumption where you could build like a Bicycle, you know, app on it, where then you can actually use that for external consumption, where customers can basically get those insights.

[00:31:29] Chapter 10: Vision and Future of Bicycle

Mirko Novakovic: So I have a question for you as a founder and CEO. Because I really like that you and also Joty, you always pick some really big problems which are hard to solve. So when you wake up, do you sometimes regret that you pick the category where I can just imagine how many technical problems you have to solve from getting the KPIs, connecting to the right system, doing the right ML, figuring out automation. I mean, it's just it's so many things. Do you just like these big problems, or do you sometimes say I should have picked building a task management tool or something like that?

Bhaskar Sunkara: No, I think I think they're fun. Also, it's always sort of like a, you know, meandering type of journey. Like when we started off, we started off pre I pre sort of like all of these LLMs. Right. So, so we initially built, you know, the data analysis part of it, which is how do you do the ML do kind of really find patterns and do all that stuff. But we have to get help from the users to set it up. Set it up the right way. So still super, super valuable, but takes much, much longer. And once we looked at how we can really harness LLMs to do it, that's when it was a compelling solution. So it's always sort of like when you get started on it, the dots start connecting. So definitely I think that's how it happened for us also. But also I feel like there is definitely a connection back to some of the things that, you know me and like also folks that you know, are, are from AppD in this team have a connection back to AppDynamics on like how do you go from technical operations to business operations? So, there's a little bit of that arc. But yeah, it's just fun to attack a big problem for sure.

Mirko Novakovic: But it's also what I understood is it's an advice and I learned it myself is just get started. Right.

Bhaskar Sunkara: Just once.

Mirko Novakovic: You get.

Bhaskar Sunkara: Started.

Mirko Novakovic: Yeah. Don't be afraid of big problems. But then maybe technology picks up, right, with the LLMs and other things, and you will figure out how to apply that to a real problem that you figure out with real customers.

Bhaskar Sunkara: Yeah. And then you create real value with, with, you know, tough problems anyway. So that's definitely. Yeah. Absolutely.

Mirko Novakovic: So what's your vision for Bicycle? What do you think? Where where can it get you?

Bhaskar Sunkara: So right now what Bicycle is, is think of it as a analyst co-pilot that sits on top of your operational data and then gives you a path from data to insight to action. Where we want Bicycle to get to. And I think we should get there pretty soon is being an AI worker for, you know, business operations, being an autonomous worker for business operations. So you plug in your bookings, checkouts, payments, chargebacks and all that stuff. And then it has the context already about who you are as a company and has all the steps of automation kind of laid out connect it with the sources of data, connect us with more of your knowledge so you can have internal documents about KPIs. You can have like, we can ingest all of that. Right? Because again, you know, the advantage is configuration is natural language. You know, you don't have to get people to configure. So just with that we you plug in your you know, technical operations. You plug in your events and we can then function as an autonomous you know, worker that basically deals both with interruptions as well as opportunities. So that's, that's that that's the vision for it.

[00:34:56] Chapter 11: Conclusion and Final Thoughts

Mirko Novakovic: Yeah. Sounds amazing. Bhaskar, thank you very much for this conversation. I really think.

Bhaskar Sunkara: You have.

Mirko Novakovic: A great vision. I, I wish that you will this time. Ring the bell. Right.

Bhaskar Sunkara: Yeah, absolutely.

Mirko Novakovic: Nobody picks you up before that?

Bhaskar Sunkara: Yeah, exactly, exactly.

Mirko Novakovic: Thanks a lot. I think it's amazing what you're building. And it's really interesting to see how you leveled up observability to the next business level and how you leverage LLMs for that. It's amazing.

Bhaskar Sunkara: Great chatting with you. Thanks for having.

Mirko Novakovic: Me. Thanks for listening. I'm always sharing new insights and insider knowledge about observability on LinkedIn. You can follow me there for more. The podcast is produced by Dash0. We make observability easy for every developer.