Navigating the Evolution of IT Observability: Observe's AI-Driven Breakthroughs and Visionary Leadership

Unlock the secrets behind the transformative world of enterprise IT observability with our esteemed guest, Jeremy CEO from Observe. As he shares his captivating journey across the pond, from the UK to the heart of the US tech industry, you'll be inspired by his experiences with tech behemoths like Oracle and Dell, and how Michael Dell's own path ignited the spark for Observe's inception. Get ready to step beyond the simple 'what' of IT issues and into the intricate 'why' as we unravel the significance of understanding the underlying causes of system failures. With Jeremy's knack for analogies, the complexities of observability platforms become as clear as day, signaling a new era where data observability solutions take center stage in tackling the challenges of modern application environments.

The conversation heats up when we discuss the cutting-edge AI tools reshaping the landscape of IT observability. Imagine AI not just as a tool but as your trusted co-pilot, guiding you through the storm of incident responses and helping you navigate complex systems with ease. We celebrate successes like Topgolf's enhanced troubleshooting capabilities while diving into how Observe leverages AI to empower users and streamline operations. Jeremy enlightens us on their incremental observability growth strategy, the symbiotic partnership with Snowflake, and the cultural blueprint that fuels their team's innovation and cohesion. Wrap up this episode with a treasure trove of insights on engineering and sales talent acquisition and the strategic vision propelling Observe into its exciting next chapter post-Series B funding.

Transcript

Future of Enterprise IT Observability

Speaker 1 0:03

Hey everybody, Fantastic topic today around the future and state of observability in the enterprise IT world with a true thought leader in the space from Observe. Jeremy, how are you?

Speaker 2 0:16

Doing great. Thanks for the invite to the show.

Speaker 1 0:20

Well, thanks for being here, and I'm really intrigued by all the good work you and your team are doing.

Speaker 2 0:34

Lots of questions, but before that, maybe you can introduce yourself a little bit about your background and the origin story at Observe. Yeah, it's a good place to start. Yeah, look, as you can probably tell from the accent, not a native of these shores but moved out in the mid-90s seems like forever ago to work. I was working at a company called Oracle, which you might have heard of in the UK.

Speaker 1 0:52

I've heard of them. I've worked for them. Yes, that sounds familiar.

Speaker 2 0:55

Yes, Well, I came out to work for a guy called Mark Benioff, which I'm sure you've heard of, and it was fairly random because I came out to learn about a product he was building in the mid 90s and I came for a week and I ended up staying six weeks and then I did the launch event up at the Moscone Center in San Francisco with him and he said, hey, you should come out and work for me. And six weeks later me and my then relatively new wife upped and moved from the UK. So that's how I landed in these shores and had a great run at Oracle, but maybe done larger tech companies over the years Veritas, semantic, most recently sort of EMC, and then we were acquired by Dell. But yeah, I always wanted to do a startup. I was actually super inspired by Michael Dell's story, even to this day. It's incredible, and that was sort of my parting words to Michael when I left Dell, which is like I really want to build something like you have, and so Observe when I joined was sort of a handful of folks, no product, maybe a little bit earlier than I was looking for, but I was really enamored with building something from scratch and the space observability I've sort of always had a soft spot for application development.

Speaker 2 2:15

Right back to the product that I worked on for Mark back in the day was a Visual Basic clone and I ran the Java development tools at Oracle. So I've always had that in me and it just struck me that modern applications are complex. When they go wrong, it can be a little bit of a goat rodeo to figure out what went wrong or why something went wrong, and it just seemed the most natural thing that everybody will need as everybody starts to build like a digital experience and you know IT, if you like, sort of is the business, not an assistance to the business, and so I was enamored with the problem domain. And then I've been on the board of Snowflake since 2015 and seen the rise of the reinvention of the database world and seen the rise of the reinvention of the database world and the founders at Observe had this idea to build Observe on top of Snowflake and that resonated big time with me because I knew the sort of architectural differences that Snowflake had and I felt like building on Snowflake could be as big a differentiator now as maybe companies like SAP and Siebel in days gone by. It was a differentiator for them to build on top of a database like Oracle.

Speaker 2 3:30

So yeah, that's sort of how I landed here. And the guy who funded the company and who was on my board I've been friends with for 20 years, ever since the Veritas days Mike Spicer. He used to work for me at Veritas many years ago and so it's great at this stage to be able to work with people A you like and B you respect. So you know he's funding the company and understands enterprise software. So that's sort of the setup.

Speaker 1 3:58

Wow, wonderful founding story. Let's dive in. I mean, folks in IT understand there are many, many traditional monitoring tools out there in the marketplace, and so how do you differentiate from those as an observability platform?

Speaker 2 4:20

Yeah, I mean, monitoring doesn't really go away. Monitoring tells you what broke, but it doesn't tell you why. And the analogy that I always think is a good one is if you, if you think about your bank account we all monitor our bank accounts and when something I don't know bad happens and you go overdrawn, you'll get an alert. Something I don't know bad happens and you go overdrawn, you'll get an alert, and so it tells you what's broken and I'm overdrawn, oh crap. But it doesn't tell you why. It doesn't give you the okay.

Speaker 2 4:55

You know, your spend this month on Ubers and DoorDash was greater than normal because your kids were back home for the weekend, and that's the reason why you're overdrawn. And so observability is about the, why you still need the, what you still need to know that something is busted, but actually you need to understand why, because modern applications, they're updated every day, and so you'll get new code in production every day, so you'll see things that you've never seen before every day potentially and so you have to be able to do the investigation, investigate that unknown problem and understand why something happened, so you can make sure that it doesn't happen again in future, which is really the job of like the SRE, or if it's a product defect, the engineering team.

Speaker 1 5:50

Wow, great illustration there. So you're sort of a Swiss army knife. You can solve so many problems or issues, but right out of the gate are there certain key challenges that data observability from Observe addresses sort of right off the bat.

Speaker 2 6:07

Yeah, I think. To just back on the investigation, I think there's like all markets that I've ever been in, you know, back in the database market 20 years ago you've got incumbents and you have new players right and the incumbents typically have designed their product for an era 5, 10, 15, some cases 20 years ago. And then you've got the newer players that you know. They start later. They can take into account the nuances of the most modern and recent environments and build for the applications of tomorrow versus the applications of the most modern and recent environments, and build for the applications of tomorrow versus the applications of the past. And so what we've seen in this space is look, it's not uncommon for folks to do some monitoring with like a Datadog, but then and that's really what made Datadog great they look at metrics, they'll allow you to set alerts and dashboards and, boom, you can see what is broken pretty quickly. But if you want to dive in and do an investigation, it's not uncommon to find Splunk in there. Or well, why is that? Well, it's more context. You've got the logs for the applications and infrastructure and that's in Splunk. And then, if you want to look at maybe the performance of the application or maybe the user experience.

Speaker 2 7:32

It's not uncommon to find like a new relic, because that's where New Relic started, and our opportunity is to say, well, look, why would you have three products when you can have one, is to say, well, look, why would you have three products when you can have one. You know, we don't really care whether we're ingesting log data or tracing data or metrics, we put it all in Observe. And so the opportunity is then, when somebody is troubleshooting, it's like, oh, look, I see a spike in that metric. Show me the spans for that metric. Okay, now show me the logs for those spans and you can do that all within one integrated product. And so that was the vision of Observe, which is, hey, you don't need multiple products, you can do it in one place. And then, as you move between these different types of data, you don't lose context because you're in the same environment and all the data is in one place.

Speaker 1 8:32

Wow, that's fantastic. Well said, Everyone likes to talk about data-driven decision-making, but at the end of the day, we're still just a lot of us making hunches. How do you empower organizations with making more data-driven decisions? What's your philosophy there?

Speaker 2 8:51

Yeah, I mean, I think it's been sort of the Nirvana state or the dream state for everyone, which is, I want to make all my decisions based off of data, and I think the challenge to date has been that people they either have gaps in their data because it was too expensive to ingest everything, or they can't make sense of it because the data is fragmented. And I'm a big believer really. I mean, everything is a matter of data. Within up data, you can solve any problem. Now you've got to do some synthesizing or some transformation of the data maybe to get the insights that you want, but it starts with being able to get the data in one place, and so the real thing that attracted me to Observe from the get-go was founders here. They didn't say hey, jeremy, we're building this great tool and my old buddy, frank Slootman's, got this great line. We both worked in tooling in our past. He was at Borland and I mentioned I did quite a stint in that at Oracle and he's got this line that tools are for fools and tools. Often it's difficult to charge good money for a tool and a tool ends up being designed for a specific purpose and when the market moves on, the tool gets left behind, right, and the thing that really I loved about the founder's approach here was they were like Jeremy.

Speaker 2 10:20

This is like we're going to build a platform to ingest streaming event data. It can be logs, it can be traces, it could be time series, it can actually be anything, and we don't care whether it's coming from Kubernetes or AWS, from your mobile app or from a robot on a factory floor. It's all just machine-generated data. And then what this platform will do is it'll transform the data into a graph to make it efficient to query. And the first use case the data is going to come from Kubernetes and AWS and distributed applications, and the workflow we're going to support the bespoke UI initially is going to be around observability and it's going to help people troubleshoot distributed applications. But we could get data from firewalls or endpoints and we could have a workflow for troubleshooting security incidents.

Speaker 2 11:15

And I guess my dream for the product is that you can build your own experience for your application and you can snap it into the product, because once you've got the data in and you've shaped the data, the UX for you to investigate that or observe that is should be relatively straightforward. And so the thing I loved about the company was the very wide angle view of observability. We're not just interested in troubleshooting distributed applications, although that is absolutely what FirstBase was. We think observability can be very wide angled and incorporate almost any component of the business. I mean from troubleshooting distributed apps to security, to understanding your AWS bill, to understanding what's going on with your robots on the factory floor. To me, that can and should all be observability and you can do it all if you've got the data and you've got the data in one place and you can got the data in one place and you can get the data in the right shape.

Speaker 1 12:22

A fascinating prospect. So Google Cloud Next is happening as we speak in San Francisco, where you are now. I hear your team is there. How do you see yourself playing in the Google Cloud ecosystem and maybe also talk about AI and machine learning, a pivotal part of your mission in incorporating data?

Advancements in AI for Observability

Speaker 2 12:41

Yeah, I mean, look, we're quite fortunate. We're quite unique in that we chose to build our product on top of a commercial database Snowflake. I mentioned earlier Snowflake. It does a lot of things for us. It's a query engine, but it also provides us the abstraction from the cloud infrastructure. So Snowflake runs on AWS, gcp, azure, so obviously Observe running on any of those cloud platforms is relatively straightforward because our interface is SQL and then Snowflake takes care of the rest. Right, and then you're right.

Speaker 2 13:17

The hot topic jour is ai um in in our space observability. There was, I'd say, like a first generation of ai technology that that was uh phrased ai ops that we were not big fans of because it was a little bit. I mean, I felt it was sort of snake oilish. It was, um, hey, we're to look at a massive unstructured data and we're going to tell you what the root cause of the problem is. And so you had a very sort of non-deterministic problem and the promise was that we can essentially pluck a golden egg out of a morass of machine-generated data. And it never really worked, like it just sort of fell flat. The new generation, you know the, the gpt based or llm based gen ai I'm actually a lot more bullish on and, and I think the the people immediately go to this nirvana state of uh, almost star trek like, hey, computer, tell me what's wrong. You know, tell me what's going to go wrong. All right, let's be predictive about that, that's. I don't think Gen AI is going to do that. Like, precision is not its strong point, right, it's a statistical engine. It can hallucinate, you know. Fortunately, you know, we're not launching missiles and doing things which can cause lives to be lost if we hallucinate. But I think that the real benefit, and I think where the win is going to be look, we have a new product, we're a new entrant.

Speaker 2 14:51

People don't know Observe or how to use Observe. If I can get a new user who's maybe familiar with Splunk or New Relic or a Datadog, if I can get them up to speed on Observe quickly, that's a win. And so the GPT-based sort of chatbot allows people to get familiar with the key workflows very, very quickly, and so we've had that in the product for about six months now. And then, secondly, we have a DSL, a domain-specific scripting language, in the product. So Splunk, you've got SPL. Observe, you've got OPAL. Well, again, nobody knows OPAL. And so again.

Speaker 2 15:27

We introduced a copilot about six months ago which allows people to specify in natural language, extract IP address from log or generate reg X for column blah, and the nice thing is is you know it's going to generate all of the opal code necessary to do that. So someone coming from a different environment, which you know we don't go really into any accounts and they're like, oh yeah, monitoring on observability, we don't have anything that does that. Like nobody says that right, so they're always moving from something, and so I'm pretty bullish on Gen AI from a like productivity of new users getting people up to speed more quickly. And the thing we're working on now is around an incident assistant. So imagine you could train an LLM on all your prior incident history and when I'm in the incident war room I can be prompted and nudged towards certain actions which may help me get to root cause more quickly.

Speaker 2 16:34

So I think there is a legit use case around essentially having a co-pilot or an assistant to help you troubleshoot an incident. I don't think it's not going to tell you what the root cause is, unless it's a problem we've seen many times before, but it can certainly. Again, I said the more junior users, rather than having antiquated runbooks that are updated by humans and get out of date. Maybe that runbook can become much more dynamic. So, yeah, I'm, as you can probably tell, I'm I'm pretty bullish on gen ai. Um, the hype is ridiculously beyond what it's capable of, um, but there's something there like, and that you know. The art of the possible, I think, is still to be defined in that space, which is why you can't ignore it.

Speaker 1 17:24

Yeah, well said, very interesting. You have lots of customers care to share any success stories where you had that operational impact. Without picking a favorite child here, can you talk about any particular cases?

Speaker 2 17:40

Yeah, I always like the very relatable stories, right, I mean Topgolf's, one of our customers. And if you're good at golf, you hit your ball towards a flag, and if you're not good at golf, you go play Angry Birds, which means you can hit the ball anywhere and still score some points on the video game simulation. You can hit the ball anywhere and still score some points on the video game simulation, and that was a classic whereby when Angry Birds crashed, the team there was scrambling to find out, like, even where are the logs for that instance of Angry Birds? Oh, wow, and you can imagine, you've got customers at a golf bay that came to hit balls and have a beer and a burger and crashed, and now you've got to move me from Bay 35 down to Bay 67. And it results in a very negative customer experience which is sort of easy to overlook as we bury ourselves in the tech.

Speaker 2 18:34

Look, as we bury ourselves in the tech. And so what we did for Topgolf is just say, look, we can extract out of your logs locations San Jose, las Vegas, so on. We can also extract out of your logs the bay number, and so now you can go like Las Vegas Bay 35, boom, here's the container logs, and that's a simple example. But we're able to add structure to something that was previously unstructured to help with the navigation through the data, so they could troubleshoot the problem more quickly. And, yeah, the net result is you've got more people playing golf instead of getting free beer waiting around for Angry Birds to be fixed. That's a great fun story.

Speaker 1 19:17

Yeah, waiting around for Angry Birds to be fixed, that's a great fun story. So for companies looking to enhance their observability capabilities in this dynamic landscape you kind of just described, I mean above and beyond the obvious, working with Observe Inc what advice would you give them to get ready for this new paradigm shift, to get their data ready for this new paradigm shift, to get their data ready for this new world?

Approach to Incremental Observability Growth

Speaker 2 19:45

Yeah, there always has to be a starting point, right. And we made a decision early on in the company's life that we didn't think it was viable to ask people to re-instrument all of their code in order to get observability. You will come across companies that say, look, if you just go and re-instrument all of your code with open telemetry statements, then we can give you great observability. That is a valid approach, by the way, but it's not practical because certainly, when you get to bigger companies, they've got years and years of legacy code that has got log statements embedded in it that they're not going to change overnight, right. And certainly the engineering teams that I've run over the years, they don't love going back and re-instrumenting code. Well, they don't like instrumenting code in the first place, but going back and re-instrumenting code is even worse, right. And so our approach has been very much like well, look, send us what you have today. This was like Topgolf we can add structure to your logs even though you haven't added structure to your logs, and that's part of the magic that Observe brings. And oh, by the way, as you move towards open telemetry which we think is a good idea we can take that too. It actually makes the job a little bit easier.

Speaker 2 21:01

But this big bang where like, oh, it didn't really work well with logs and a few metrics, let's throw that away and spend the next year re-instrumenting with open telemetry, I'd be like, yeah, that's not going to fly If you're a tools team trying to convince the engineering team to go do that is not going to go well. And so what we tend to do is show people what observability could bring, and then we tend to say, well, look, let's take your log environment and let's make that better. You'll probably save a little bit of money as well, because we've got a very different architecture. And then after that, maybe we pull in the time series data and the metrics. And then after that, maybe we pull in the time series data and the metrics and then after that maybe we pull in the tracing data. Maybe you want to go to tracing first, or maybe you haven't got tracing at all, that's OK.

Speaker 2 21:47

So the sort of path to like full observability we tend to pick off one particular workload or application. First we tend to start with logs, and then we go to other types of telemetry and then go to other workloads. That tends to be the process. So yeah, like any of these big shifts, don't go big bang, don't bite it off all at once. Start small, get some wins, demonstrate the value, and if you can't demonstrate the value, don't go any further. But if you can, then keep going. You know, that's, that's sort of be my, my advice to folks.

Speaker 1 22:26

Yeah, well said Crawl walk and then run. Always good advice. Tell us a bit about the culture of the team that you've built and are directing. Maybe give us a peek behind the curtain. Where are they and are you guys distributed, or do you have a headquarters? What's your philosophy?

Speaker 2 22:44

there fairly centralized from an engineering standpoint, here in in san mateo. Um, we do have some remote folks, but I'd say the you know, probably 80, 80 plus 85 is is here in san mateo. Um, yeah and uh. I mean, I think, like a lot of companies we've, we've got like a hybrid work environment. People come in three days a week. Um, we started initially with folks just figuring out which days they wanted to come in and I think more recently we've thought, well, okay, it's probably more productive if we just say Tuesday, wednesday, thursday in the office, monday, friday come in if you want, if you don't, that's fine too. So to me that seems to be striking the right balance between giving people the flexibility that maybe they've become accustomed to and and as I think it's still very important and it's very important to me like building a team and a culture um, it's still hard to do on zoom, um, so there's no substitute for getting the team together. And I say to folks look, you might actually get less work done if you come in the office. I'm okay with that, because we get the intangible benefit of human interaction and building trust and building a team and building a culture, and those things become very, very important.

Speaker 2 24:06

As you start to scale the company right, when you start hiring lots of people, you have to have a definite way of working, you have to have a definite culture. And you start to scale the company right, when you start hiring lots of people, you have to have a definite way of working. You have to have a definite culture and you want to pass that culture on to new folks. And I feel like you know grads and interns. We insist that they're in the whole time. I know what I was like when I was a grad. It's not to say that you don't have incredibly disciplined grads and interns that are capable of staying on point and on track five days a week, but I was never like that. So I feel like we owe it to a lot of the newer folks coming into the workplace, getting them into good habits and a little bit of structure and having a coach or a mentor for them to get going. So that's sort of the inside view of where we're at.

Speaker 1 24:54

Love that, really really fun. What are you looking forward to over the next weeks, couple few months? Event-wise, product-wise, travel-wise, what's on your mind?

Speaker 2 25:07

Yeah, well, we just closed out our Series B round, which is good. I'm personally looking forward to not having to raise money, so that that's a big win. So we're going to focus on spending money versus raising money. Um, and we're having a pretty good start to the year, so there's certainly still hiring that we're doing, um, a little bit on engineering, but more on the on the sales side of house as we build the capacity for next year. I think that announcement I think for startups, funding announcements tend to be your biggest opportunity of getting some airtime and getting the word out there. So, as I was saying earlier, it's going great at Google Next and I suspect that's because we've been in the press. So I'm looking forward as we sort of go through a lot of the DevOps and SRE shows that we frequent. Hopefully there's a little bit more interest this year and a little bit more business on the back of it.

Speaker 2 26:02

We have a big partnership with Snowflake. I mean, as I mentioned, we're an application built on top of Snowflake and we developed back in November something called a connected app in Snowflake speak and what that means is that Observe, instead of the customer running on Observe's multi-tenant environment, observe will run against the customer's Snowflake tenant. Observe will run against the customer's Snowflake tenant. So you know, if they spend a bunch of money with Snowflake, they can leverage that spend to run Observe. And so I'm actually pretty excited about the Snowflake Summit this year, which for the first time is in San Francisco, and we've got this connected app that we built last November, and so I'm pretty excited to see what the interest is like from the sort of larger enterprises, which Snowflake knows a whole lot better than we do.

Speaker 1 27:00

Well, it's exciting times and great to see events back in San Francisco. If you're at Google Cloud, stop by say hello to Observe. And yeah, thanks for joining me, sharing your insights and experience. It's quite informative, jeremy. Thanks so much.

Speaker 2 27:11

Yeah, thanks for the invite, Great to be here.

Speaker 1 27:13

Thanks everyone for watching, thanks for sharing and resharing as always, and stay safe.