Siddhartha Ahluwalia 1:04
Hi, this is Siddhartha Ahluwalia. Welcome to The Neon Show. I’m your host and also managing partner at Neon Fund, a fund that is invested in some of the best enterprise AI companies between US and India Corridor, like Atomicwork, CloudSEK and SpotDraft.

Today I have with me Manish, President at Arize. Manish, welcome to The Neon Show.

Manish Jindal 1:23
Yeah, good to be here. Thanks for having me.

Siddhartha Ahluwalia 1:24
Your journey is, you know, amazing. You have scaled Cloudflare, one of the employee number 45, stayed there for 10 years.

Manish Jindal 1:33
That’s right, yeah.

Siddhartha Ahluwalia 1:34
And you saw the company from, let’s say, go from 200 million in valuation to public market cap of $60 billion. That’s an amazing journey. Is that where you learned the chops of GTM?

Manish Jindal 1:46
Yeah, absolutely. I mean, my background is in go-to-market. So I did, I think first company I learned go-to-market was at Dell.

But I would say that Cloudflare was the one where I really applied what I learned in the past and continue to hone it. And also kind of as the company starts to grow, the way you go to go-to-market changes as well. So I think Cloudflare, I would say, is where I learned the most from my experience perspective.

Siddhartha Ahluwalia 2:13
And your choices of companies have been great, right? You were early member in Splunk, Cloudflare and Arize. How did you make those choices?

Manish Jindal 2:21
Yeah, yeah. I mean, for me, I think sometimes luck plays a role. But I think broadly, when I think about the companies to join, there are a couple of things I always look for.

One is, I’m a big believer that if the market you are in, if the market has a big opportunity, you as a company will get a lot more chances to make it work, right? If you’re in a market with a small TAM, you do one thing wrong and you’re out of business more than like, right? So for me, when I chose Splunk, I just felt that big data was something up and coming.

And I felt that everything up till that point was structured data indexing. And Splunk was the first company where you can index unstructured data and make it searchable. And I just felt that the TAM of this is massive and Splunk was doing great work.

Similarly, for Cloudflare, the initial market of Cloudflare was in the tens of billions of dollars. And Cloudflare actually, I felt, found this very niche market where anybody on the internet, who doesn’t have the scale like the larger companies, you know, like think about the Yahoo’s or the Google’s of the world, but they still need their website to be fast. They need the website to be secure.

They need their website to be online. And there was nobody serving that. And I just felt that the market is big.

And it’s just underrepresented market. And there’s a company actually who’s thinking about that unique opportunity. They should be, should do well.

So I think it’s just always for me come down to how big the opportunity is. And then I always bet on the team as well. You know, sometimes, for the most part, like, I mean, you pick any company, like a lot of multiple people have the same ideas.

And very few founders are able to make it successful. And it really comes down to how, what is the belief of that founder in that company, right, or in that market. And usually that is the belief which carries you through the, you know, the ups and downs of a company.

And I felt that, you know, the Splunk founders were like really believed in what they were doing. Same thing I felt about Matthew and Michelle. They genuinely believed in what they were doing.

And I could see that they would do whatever it takes to make this work. So, and same thing, you know, now I joined Arize. I felt the same way, that big market and the founders really believe in the company and the market.

So luckily for me, it worked out.

Siddhartha Ahluwalia 4:36
And you have always chosen companies on the dev tooling side rather than application side. Why is that?

Manish Jindal 4:43
I think that’s by chance it happened. But I think only thing if I think about my background, so my background is in databases. So my first job out of college was, I was a backend developer.

Siddhartha Ahluwalia 4:53
Maybe you can share a little bit about where did you graduate from?

Manish Jindal 4:56
Yeah, absolutely. So I studied mechanical engineering from Punjab Engineering College, in 2001. And my first job was a software engineer backend developer for databases.

So I knew, I very early learned how to write SQLs and Oracle database and whatnot. So data is something I was always comfortable with. And I understand what that really is.

And, and that’s why, you know, Splunk kind of, I understood what this Splunk is trying to do.

Siddhartha Ahluwalia 5:21
And how did you move to the US?

Manish Jindal 5:23
Yeah, so I came to the US to do my master’s. So I came to the US in 2003 to study master’s in Management Information Systems, MIS, at Texas A&M University. So I came for that, did my master’s there.

And then I got a job at Dell in Austin, in the initially in the business intelligence group. And then I moved because I always had a knack for the business. So I then moved into the marketing team.

And then I also ended up doing MBA from University of Texas in Austin. So I ended up doing two degrees in the US. But yeah, that’s kind of my journey has been.

Siddhartha Ahluwalia 6:03
Got it.
And how was the breakthrough in Splunk? Like, how did you identify first of all, because dev tooling was not popular back then?

Manish Jindal 6:11
Right, right.
So for me, how this happened was, it was one of one of my friend from Dell was working at Splunk. And so after going to business school, I actually ended up going to management consulting, you know, like they have they were at that time was the highest paying, highest paying employers, right? I went there, I did two years in the management consulting, and I was in the Bay Area.

So I moved to San Francisco after doing my MBA. And, and that was the time when, you know, the companies like Facebook and, and Uber and Airbnb was just coming up a little bit. And, and, and, and I remember, Facebook was the company which went public when I was here, I think around 2009, 2010.

And, and, and I felt that, you know, this is where I want to be, you know, I want to be part of this young, some of these younger companies. And if the company does well, one is I’m going to have amazing experiences. And that would be a way to really create generational wealth, you know, is part of riding that wave.

So I started applying to a lot of startup startups at that time. But startups usually do not hire people with like MBAs and who come from management consulting for their in their mind, you guys, you know, all about process and data and analysis. But as a startup, you do none of that, right?

A startup is all about taking actions and, and learn from that experience and, you know, improve from there. So for me, Splunk happened because I knew this person from Dell, he was working at Splunk and he made the introduction and, and I was able to get in through that referral. And that’s kind of how Splunk happened.

But I, I got excited about Splunk because I could understand what the product was, and it was making a lot of sense. But I do remember at that time thinking, like, I think I want to go work for Airbnb, because you know, there’s always, you know, these sexy companies out there where you like, I think that is the cool company, but I could not get an intro at Airbnb or Uber or Facebook for that matter. So for me, the company which was in data infrastructure kind of chose me and I, and I chose them.

And that’s how kind of it happened.

Siddhartha Ahluwalia 8:13
And why did you leave Splunk early?

Manish Jindal 8:15
So I left Splunk early for, for a reason that I just did not get along well with my boss, basically, and not, not like in a personal sense, but I think I just did not agree for the decision making. And, and I, and I just felt that, that, you know, the decisions being made were not right, in the right interest of the company. And I just could not take it basically.

And for me, what happened was there was this other company which was doing really well, it’s called Lattice Engine. They were in, in the, in the data space. So what they were doing was using predictive analytics to predict what should be your upsell, cross sell place, you know, like think about, you know, when you’re on Amazon, it tells you if you buy this, you should be buying this.

They were doing that for B2B companies. And I felt that that was an interesting space. And they were leveraging the data to make those pieces.

And I just felt that I can really, learning from Splunk and apply to this company and, and, and I just felt that this was a better fit for me. But on the hindsight, I, after leaving Splunk, my actually boss was, got fired within two months of leaving. So it became a moot point, like in a way moot point, but this is what happened.

So, but I learned from that experience is like never leave a company you believe in for, for like, you know, these kinds of reasons, you know, you have to stick it out because things change really fast. And one thing doesn’t change is like, you know, the, the, the trajectory of the company, like if you are an interest company, which has amazing product, customers are happy, big TAM, stick it out, you know, you will, those problems will go away over time. And I, those things stayed with me at Cloudflare and I ended up staying there 10 years because of that.

Siddhartha Ahluwalia 9:46
So, so you are sharing some interesting trend, right? A company which has a large TAM, right? The founders are directionally right in how they’re building the product, because at early stage, nobody can tell which direction the company can go.

But if it’s a large TAM, you give yourself enough surface area.

Manish Jindal 10:06
Surface area to, to experiment and fail and, and still being able to recover and put something, build something from there.

Siddhartha Ahluwalia 10:13
And let’s say some of the other dev tooling companies, or for example, companies, which are infracompanies that, that I believe you have always been part of infracompanies. And now they are becoming more and more cool because, you know, they say in Gold Rush, give picks and shows. Is that what you are always looking for, even when you are investing as an angel?

Manish Jindal 10:35
Right.
So I think, you know, I believe that, you know, like I call it infracompany, I sometimes call it plumbing companies. You know, your plumbing is something when you’re building a house, right? Once you put the plumbing, you never change it.

That’s the reality of it. Right. And unless, even if the plumbing breaks, you just go whatever plumbing you have, you fix it.

And I feel that a lot of the infracompanies are bad, where if you can be part of, part of the journey for, for a company, you know, and be part, be the plumbing for them, they will just never replace you. So, so that’s one of the reasons I enjoy Infra. I just like Infra.

It’s not sexy, right? It can, you know, it does not have, does not get the same level of publicity, like as the application companies. But I just feel that, you know, sometimes it’s the boarding industry, which are the one who end up doing well in the long run.

Siddhartha Ahluwalia 11:26
Yeah. But today, if we see, you know, companies that are getting love and attention, data breaks, valued almost like a $200 billion company, again, a plumbing company, you don’t change your data store.

Manish Jindal 11:40
Right, right, right.
Yeah. So those companies came out, but I think if you think about it today, right, I mean, if you look at the world today, a lot of the AI application companies are growing way faster than the plumbing companies, like pick like Replit or pick Lovable or pick, you know, any of these companies, they are able to go from like 10 million to like 100 million within like a matter of months or quarters, right, which is not going to happen for plumbing companies, like on Infra company, because Infra companies usually grow slowly, right? I mean, typically, historically, and same if you look at the database journey, they’ve been now around since 2016, you know, it’s from the Spark days.

And now they’ve been around 11 years, and it took them a while to get to where they are. So I think it’s a, in the long run, if you have a great product, like I just said, and you are the plumbing company, Infra company, you’re going to do well. But it will take you a while to get there.

Typically, it’s not like, you know, overnight, you will not become a success, you know, like if you’re a database company or something. But today, you’re seeing like a lot of the companies become overnight success with application AI companies.

Siddhartha Ahluwalia 12:42
And which are the spaces that you think will grow in the next 10 years, for example, market? Before starting a podcast, we were discussing about voice. So in your opinion, including voice, what are some other spaces that you see real revenue growth, not just factual valuation growth?

Manish Jindal 12:58
Right, like, and you mean the from a AI application perspective?

Siddhartha Ahluwalia 13:02
In general, but carrying that tailwind of AI?

Manish Jindal 13:06
Right, right.

I mean, I think I do believe that the middleware companies, right, which kind of sit between the application and the models should do really well in the next 10 years.

Siddhartha Ahluwalia 13:18
Example?

Manish Jindal 13:19
So like, think about like Arize, I picked, they are in the evals and observability space, where the biggest challenge today, you know, as you think about building the AI agents is, is how do you know agents is going to do what it’s supposed to do? It’s very easy to build an agent, but it’s very hard to scale an agent.

So and you need the right infrastructure for that to happen. And those like, like evals is one example where it tells you whether your agent is doing what he’s not doing observability is, you know, make sure that your agents are performing as they’re supposed to in production. I think that will be a big market.

You know, it’s very similar to like Datadog example, what Datadog did 10 years ago, or Splunk did to 15 years ago, or like all of that. So I believe that this particular space, and if any space should be like hundreds of billions of dollars, right? I mean, and the reason why I feel it’d be that big is if you add up market capital, like Datadog, Splunk New Relics of the world, it’s more than 100 billion today.

And I think AI observability is going to be bigger than, you know, the infrastructure observability or the APM today, right? So it should be like a big, big opportunity in my mind, like, you know, that’s definitely it should be big. I think inference should be another big one.

You know, as you think about that, that potentially could be a big market. I mean, I think anybody who’s going to like sit in the middle, I think about, you know, when I think about the DevOps, the contrast with DevOps layer, right? You know, DevOps, you have your observability, you have your security.

I think, you know, they potentially is like security companies should continue to do really, really well. You know, if you look at Cloudflare today, I mean, they are doing really well, because now people, you know, believe that they’re going to play a major role in the world of AI. So I think it’s the boarding industry is going to do well.

The application side should do well. But I think the problem, like the way I see it is, there may be 20 companies will do well today, but off that maybe one or two will be successful in the long run, because it’s very easy to challenge the mode of those companies, right? Because end of the day, their biggest mode is their consumer, right?

Who’s using those. And these days, we know that consumer is very finicky, they don’t mind moving from one application to other like, you know, in a given day, like, if you look at the developers or the users, they have a very low attention span.

Siddhartha Ahluwalia 15:38
Yeah, they immediately moved from cursor to cloud.

Manish Jindal 15:41
Right. So think about cursor and cloud, right? I mean, within a day, you people make the switch.

And same thing, I feel like, you know, if you’re Lovable or Replit, people will move very, very fast, right? So that’s why, even though the TAM wise application maybe have a bigger TAM over the next 10 years, but it’s very hard to pick the winner in that space. It’s much easier in my mind to pick winners in more of the middle, the infrastructure space.

Siddhartha Ahluwalia 16:06
And why do you say it’s easier to pick winners in the middle Layer?

Manish Jindal 16:08
The plumbing, right? Once the plumbing is in, it’s very hard to change.

Siddhartha Ahluwalia 16:11
But in application, usually it happens like winner takes all, in plumbing also, it’s like the winner takes all market.

Manish Jindal 16:17
Not always. I mean, pick any industry, right? Like, even if you pick observability as an example, you have, what, six or seven companies who are more than a billion dollars today.

Similarly, if you look at the databases side, you have multiple companies who are doing really well. So I think it’s not that winner takes it all, especially on the infrastructure side. I mean, if you look at the cloud for the space, which is the application security and performance, there’s still multiple players.

There’s Akamai, there’s Cloudflare, there is Fastly. All are billion dollar companies, right? So there is a space for multiple companies, I believe.

But in the application side, I think what’s happening is that the user actually kind of move in hordes. They love one company today and they all are using it. And then next day they start using other, like other company and they just, the whole move from one to other.

And they don’t really have no reason to stay back. If you look at the history, right, the companies which own the data did really well. Like think about Salesforce, right?

I mean, Salesforce did really well as a company because they now have all your business data, all your sales data. And it’s very hard to move the data from one platform to other platform. And also, you end up building all these workflows and process on top of it.

So it becomes extremely hard to move. And that’s why Salesforce is still a major company. But if you think about a lot of the companies which were not owning the data and a lot of workflows company came, they kind of came and went and did not really become that big.

So I just feel that in the long run, it’s much easier, I believe, to pick a winner where you’re sitting on the back end versus on the front end, unless you own the data like Salesforce.

Siddhartha Ahluwalia 18:04
But what about enterprise applications? Like the examples that you took, Lovable, Replit, they are more, let’s say, prosumer kind of application. What about enterprise?

Manish Jindal 18:16
If you are enterprise application, I believe that if you are the one who keeps the data, then you will be successful because moving data is very, very hard unless that becomes of the future. So I still believe that any enterprise, like if I’m Workday or I’m Salesforce or HubSpot or Freshworks, I mean, you have a strong place still because you own the data.

Siddhartha Ahluwalia 18:38
But let’s say today HubSpot is 80% down. So how do you justify it?

Manish Jindal 18:44
I mean, that’s a pivotal shift in the market. Because I think now what’s happening is that the software is potentially going away and getting replaced by AI. If you think about the world where in the future, I believe that you are not going to do clicks to get something done.

You’re just going to ask in a natural language. So which means that I think these earlier software companies more than likely is going to get replaced by AI systems where their AI system is not going to have any UI or none of that, no clicks. And it’s just basically interacting in a natural language.

And we see this now with the cloud code or with the cursor and all, it can do wonders, right? I mean, like even in my current company, all our systems data like Salesforce, HubSpot and everything, even Gong data, we actually put into BigQuery, every data. And then we have cursor sitting on top of BigQuery.

And if anything I want, I just ask cursor and cursor give it to me. So I don’t need any analysts. I don’t need anybody to stitch the data together.

None of that. You can just ask in a natural language. So I think in the future now would be as a user of Salesforce, they’re not going to go clicks and add information.

I think this is going to tell the system, I’m working on this deal, open this opportunity, do this stuff. So unless these companies find a way to move away from software and become AI companies, I feel that they’re not going to be very big in the future.

Siddhartha Ahluwalia 20:17
And, you know, looking back at your Cloudflare journey, what was the revenue of Cloudflare when you joined?

Manish Jindal 20:24
I mean, when I joined, there was very little revenue. I think it was like a few millions here and there.

Siddhartha Ahluwalia 20:30
A single digit million?

Manish Jindal 20:31
Single digit million, for sure. So what we had when I joined the company was a free plan and a $20 plan, which we called ProPlan.

Siddhartha Ahluwalia 20:42
There was no enterprise plan.

Manish Jindal 20:43
There was no enterprise plan. So when I was hired, along with Chris Merritt, Chris Merritt was the president of the company. We were hired to help build out basically the sales motion because the PLG motion started to do well.

And we were having, we ever had a lot of free users and good amount of pro users. And we felt that it was the right time to start thinking about building the enterprise motion or the sales motion, not enterprise. And that’s kind of when I joined, basically.

Siddhartha Ahluwalia 21:11
And when you exited, what was the revenue?

Manish Jindal 21:12
When I left, it was like about 1.2 billion revenue.

Siddhartha Ahluwalia 21:15
Today, it will be like 2 to 3 billion?

Manish Jindal 21:17
Today, I think it’s a little over 2 billion now. Yeah. So it’s growing very fast. I mean, I left the company end of 23 and they are growing somewhere between 30 to 40% a year. So which means should be over $2 billion now. So yeah, this is a strong business, it’s a good business.

Siddhartha Ahluwalia 21:34
So if you can share your playbook of some of the mistakes that you did in building the GTM at Cloudflare and what other things that you did right?

Manish Jindal 21:43
Yeah.
So what things we did right and mistakes we made. Okay. I mean, what we did right, I think, and as I think about our Cloudflare journey, what things we did right was we never tried to boil the ocean.

Actually, I’m a big believer that you can’t be good at everything. And what I mean by that is when you’re building a company, you can’t be good at serving all kinds of customers. You can’t be good at serving a customer through a self-serve motion and also through a sales team.

You can’t be good at selling a customer in the US and also in Europe and also in Asia. You can’t be good at selling like 2 products at the same time, right? So you have to pick and choose a combination which really works for you well.

And you have to pick the right customer segment, you have to pick the right motion, you want to sell it through and they’re in which market. And at Cloudflare, I felt that we had that understanding and clarity that over the long run, this company has a massive potential. But over a short run, we just have to go win a particular segment of the customer.

And if we do a good job with that, we will underwrite to go win the other segments of the customer. So we’d never tried to boil the ocean. So initially, our focus was very much bring a lot of the customers to the PLG and get them to value really fast.

One of the obsession at Cloudflare we had was how to provide the value to the customer really, really fast. So our self-serve motion was built in a way that you come, you swipe your card, you can turn on the service within a matter of minutes. All you have to do is you move your DNS to us and suddenly your website is going to load faster, it’s going to be secure, it’s going to be reliable, all of that.

So the holy shit moment we talk about would happen really fast with Cloudflare. And what that allowed us to do was just build this flywheel of a lot of developers, a lot of the webmasters and founders just start putting their websites on Cloudflare. And we felt that we had this very unique motion as a company.

And this is something we should build on that versus start, go find a new motion. So initial enterprise plan was nothing but the same product which you have on PLG with the white glove service. Because what happens with a lot of customers is they come through PLG because they love that it’s very simple to start, very simple to use, very simple to get to value.

But they want something more than that. Like they want to make sure that, hey, there is actually somebody who can fine-tune my environment to make sure I have configured Cloudflare properly. Or if something breaks, actually, there’s a phone number I can call.

They also like, hey, I don’t want to have to pay the credit card every month because then I have to go expense. It’s just too much work. I’d rather pay on an annual basis.

And we started seeing those customers, right? So first enterprise customer we have was Bain Capital. So Bain Capital was using our Salesforce product and they were happy with it.

And we just put a phone number on our website if somebody wants to call us. And that phone number was pretty much like cell phone, my cell phone. And we had a phone basically on the desk where I was sitting and Matthew or Michelle used to sit, Chris used to sit.

And the phone would ring and one of us will just pick. And the idea was that, hey, if somebody calls, we can learn something from it. And we actually got a call from the CIO of Bain Capital.

And he started, he was like, hey, I love your product and all the good stuff. And but he’s like, but I need more than what I’m getting from it. And what he was looking for a white glove service.

He’s like, I want, I was like, I’m not confident that we have configured this properly. I want to make sure that there’s a solution engineer who can help me with it. And then he’s like, I just hate two expenses every month.

You know, what are you doing? And I remember asking him, how much would you like to pay if he can, you know, put all this together into as a plan for you? He’s like, yeah, something around $3-4,000 a month is okay.

And I’m like, okay, that’s like almost $50,000 a year. And I was like, you got it. And that’s kind of was the day we like, okay, we are launching our enterprise plan.

And it was nothing but our PLG product with all the vital service and the things people wanted. But over the period, then we end up adding a lot more products and pieces, which made the enterprise very unique compared to the software. But the initial couple of years, the enterprise product wasn’t like different than the PLG.

It was nothing but a white glove service, right? So this is kind of how Cloudflare go-to-market was built, was in a very thoughtful way without trying to boil the ocean, really be fine tuning to the needs of the customer and the serving that customer really, really well. And then we just continue to like add more and more layers of go-to-market to it, right?

Then initially, it was very much bringing PLG, selling them to enterprise. Then we started also going with outbound motion to sell, get customers who are not coming to the PLG. Then we went and opened an office in London.

Then we went and opened an office in Singapore. Then we will launch our second product. Then the motion become from acquisition to like more expansion.

And then we felt we are ready for the large enterprises. Then we really went after the big companies. But it was like in a very thoughtful, sequential way we went about it.

And one thing was very clear to us that we wanted to grow fast, but we never wanted to like that we have to triple the revenue every year. We were like, if we can double the revenue every year and go slightly faster than that, I think we will build a very strong business which will survive the test of time.

Siddhartha Ahluwalia 27:06
So if you can recall, roughly, which year did you cross? 50 million and which year you cross 100 million?

Manish Jindal 27:12
Yeah, yeah.
So I believe we crossed 50 million. So I started with Cloudflare in 2014, early 14. I think we crossed 50 million in, I think, end of 2016, in three years.

And then at that time, we were more than doubling the business. So then we must have crossed 100 like within the next nine months. So we went public in 2019.

And at that time, we were doing about 300 million. So in our case, I think we went from like a couple of million to 300 million in a matter of five years. And then after going public, we actually accelerated our growth because as a public company, we were able to have bigger awareness and then COVID happened.

And then we went from 300 million in 2019 to a billion in 2023, basically. When you basically… When the end of which I left, basically, right.

So the growth happened fast. But it was not like what you see today, where the company is going from 10 billion to $100 million in a year, going from 100 to 150 million in a quarter. It was not like that.

It was a little more organic growth. But we were always close to doubling the business in all the years. Or in some cases, tripling the business.

Siddhartha Ahluwalia 28:36
So what I’m learning from you is today, like most of the founders, they rush into their enterprise and go to market really early on. But you took your own sweet time to build that enterprise portion.

Manish Jindal 28:48
Right, right. Because, you know, especially if you’re an infrastructure product, right. The number one thing customers want from you is the stability of your net, right? Like, are you stable? Are you going to not go crash? Whatever, you know, like, you know, I’m building.

So if you become that critical point for a customer for their application, you have to be very careful. Can you really sell them well or not? You know, end of the day, you only are successful if your customers are successful.

You know, no point signing like a very fancy customer. But if you can’t sell them well, and more than likely, you know, they’re going to churn. So you have to be intellectually honest and say, based on where you are, what kind of customer can you sell well?

And you do a hell of a job in making those customers successful. And then at the same time, continue to build your capabilities. And then when you feel that you’re ready to serve, like the biggest indicators of the customer, go do it at that time.

Other challenge you could also happen is if you go bring a very large customer very early on, then that customer is going to force you product in a direction you may not want to go. Right. I mean, if you bring in, I don’t know, Apple as a customer, more than likely, like Apple is going to dictate your roadmap.

So one of our competitor, Akamai, biggest customer was Apple. And we would hear all the time is Akamai roadmap is dictated by Apple. And so for us, it was very clear that we don’t want a customer like Apple and all early on, because for us, the mission of the company was how do you make internet, help make internet better for everybody, which means that we wanted to serve a wide range of customer, like wide range of the internet versus the select few. So we intentionally actually would not go after the large customers.

Siddhartha Ahluwalia 30:28
So you said no to Fortune 500.

Manish Jindal 30:31
Like 100 percent, right? If they come to us, we may entertain, but we never went to them. Like in the first three, four years, we did not go.

We did not go after any of the Akamai customers at all. But if they come to us, we will. But there are times where we did say no to the customers, like, no, we are not ready for you.

And I think that honesty is important because at the end of the day, the great businesses are built over decades, not months or years. And you have to have to take that longer preview is like, how do you want to sequence things? At the of the day, you have to know no customer is too small. I mean, if you’re happy customers is what you really need because that really build a flywheel for your next set of customers. So you are better off doing that versus getting the first million dollar customer.

Siddhartha Ahluwalia 31:14
But founders, how do they overcome that fear that if Apple is knocking on your door and you’re not ready for that to scale yet, Apple might go to your competitor.

Manish Jindal 31:24
Right. Which is, I mean, yeah, you’d let them go, right? I mean, because the thing about Apple and all is that they do come knocking the door every two, three years about a particular tool.

Right. So it’s not that if you lose Apple today, you completely lost your right to Apple. You will get another chance, you know, like in the next couple of years.

And that’s OK. So that’s where, you know, that patience becomes very important, where you actually intentionally say, I’m not ready for you, but, you know, let me come back to you in two years and then you will win the business. Like Apple became a Cloudflare customer eventually.

And so it’s not that Apple did not, but it did. But that happened like in 2020 or something like that.

Siddhartha Ahluwalia 32:08
After almost like six..

Manish Jindal 32:10
Six, seven years. Yeah. Yeah.

Siddhartha Ahluwalia 32:13
So among all the companies that you joined, like Splunk, Cloudflare, Arize, they had, I would say, very early signs of product market fit at the time you joined. How did you decide that the PMF is there or not?

Manish Jindal 32:29
Yeah. I mean, I think because I had a choice to pick a company with the PMF, then I did choose a company with the PMF. Right.

It makes your life much easier. You know, PMF is something you can’t force it.

Siddhartha Ahluwalia 32:39
How do you define PMF in your definition?

Manish Jindal 32:42
Yeah. I mean, the PMF in my mind is like there’s a pain in the market and you clearly sold that pain for that. You know, that’s a PMF.

Right. Because a lot of times the pain is not clear, but you have built a product and you continue hunt for the pain. Right.

So it has to be. So for Cloudflare, for example, the pain was real in the sense that like, let’s take Splunk. Right.

The pain was real. There was the massive amount of unstructured data out there, but it was not indexable before Splunk. Right.

It’s not, if you can’t index the data, you may not, can’t make it searchable means that the data is useless. Right. But there was so much unstructured data that there got to be a use case for it.

And Splunk did have that problem early on where they were like, Hey, we are the big data platform for unstructured data. And a lot of customers like, I get it, but I don’t know what do I do with this data. And that’s when the Splunk actually thought about, let’s build a use case of which we are done.

You know, we can be, we can be powerful. And then they, that’s where they got into that overall, the, you know, the data center observability space basically. Right.

Which is idea is that there’s insane amount of logs are not being created. If you can make them searchable, you know what the hell is going on. And so do you have, now you build a use case and that use case is very clear.

There’s a real pain. Everybody who has the data centers, they have the bunch of switches and the routers and whatnot, and they need to know are they working properly, if something’s going to fail or not. And now you have, you’re solving the real pain.

So I joined at the time when they kind of figured that out. It was clear, right. That, you know, you need this and it’s And for Cloudflare case, it was very clear, you know, like it’s very clear that every website, like I remember reading this, this study about Google because Google will track, you know, think about their ad revenue.

Right. And their ad revenue was directly correlated with how fast the website will load. Right.

You know, which is, you know, if the website is fast loading means the customer is going to take action, they’re going to buy stuff and, you know, it’s going to generate more ad revenue for them. And, and, but if you look at, you know, back in 2012, 13, there was only 2% of the website who actually was using some kind of a CDN solution or some DDoS or whatnot. And, and, and I’m thinking like, all these websites do need this solution, but there’s nobody who’s, who’s delivering it except the Cloudflare.

So the PMF was very clear. Like there is a PMF out there for what Cloudflare is doing. And same thing I saw with, with Arize where the traditional observability just doesn’t work for AI.

And, and everybody, it’s like every company you talk to, they have either working on an AI project or they have planning to work on AI project in the next 12 months. And, and, but these, none of these companies know how do you take this agent at scale? And one of the key pieces to run agent at scale is to have the right tooling from an evaluation observability perspective.

And there are only very few companies who are actually doing a good job of that. So the PMF is again, very, very strong here. So for that’s kind of how I look at it is like there’s a real pain in the market and you actually do sell, solve that pain very, very clearly, then you have a PMF.

Siddhartha Ahluwalia 35:43
And in case of Arize, because now you are well known in the Bay Area ecosystem, you know, as a leader who could take a company from let’s say 10 to a billion, you would have multiple offers, I assume, right? So, so why only this space? Because now you are not only looking at, I assume companies with PMF or infra companies, but why only choosing eval and observability for AI?

Manish Jindal 36:08
Right, right.
I mean, I think, partly because of familiarity. So before joining Arize, I, I was with Insight Partners.

Siddhartha Ahluwalia 36:19
You worked full time with?

Manish Jindal 36:20
I did work full time with Insight Partners. And Insight Partners had two companies, Fiddler and Weights & Biases, which were in the similar space. So by being there, I understood this industry well, and I understood the challenges these companies have, and from a product perspective, what do they lack?

And I felt that Arize have everything these companies don’t have. And even though without these companies didn’t have a lot of these things, but they were still doing really well. Just that just kind of gave me the conviction that Arize have a very strong future, basically.

You know, end of the day, like we all have a little bit less options, but you know, I always feel that if I know something, I feel familiar about something and comfortable with something, I’ll just do it. I will not chase the shinier thing than that. You know, I’m a big believer of that.

I usually don’t chase the shiniest thing.

Siddhartha Ahluwalia 37:10
No, you don’t super optimize.

Manish Jindal 37:11
I just don’t optimize, you know, like I calculate the odds of something and if something has a good odds, then you just go do it.

And for me, Arize felt that, one is, its space was interesting. The product was amazing. Like, you know, Arize customers love their product.

Like no matter who you ask, they just love it. You know, I was just in India some time ago and Arize has this open source protocol, Phoenix. And I can’t tell you developers like, I love Phoenix.

It’s a game changer product. So for me, I felt the product is great. The founders are technical, super smart, bright, and they, I would bet on them any given day against any other founder, by the strong, strong founders and strong product leader visionaries, right?

And I felt the founders are strong. The install base of Arize was very, very good. So Arize have customers like Atlassian, Walmart, Wells Fargo.

They have like Uber, DoorDash. They have bookings.com. You know, like when you look at this customer, you like these, you can’t make these customers happy unless you have an amazing product.

So I build the conviction that the product is amazing. Space I understand and the founders I love, like what more do I need? Basically, you know, like I could keep chasing the shinier thing, but then, you know, then I was like, no, this is a really good product, really good feature.

And I just felt that this is a fit within the wheelhouse of the work I’ve done in the past. You know, I like being in a, like a dev-centric company where, you know, you have the developer love and this company has all of that. So I chose it. Yeah.

Siddhartha Ahluwalia 38:44
Then probably I assume, you know, you look at things like 10-year stints. So this is probably your last 10-year stint.

Manish Jindal 38:53
Yeah. I always look at like things in decades, not in years. I mean, that’s why I’m a genuine believer in that. Greatness doesn’t happen over in a few years. I mean, sometimes you get lucky, but usually it doesn’t happen.

Siddhartha Ahluwalia 39:04
And why join a company whether now you had like large credibility of you could have started your own company?

Manish Jindal 39:10
Right. Right.
Yeah. I mean, I think it’s just, I could have, but I just never, you know, I’m not very technical. I mean, even though I have an engineering background, right?

So one is I need a technical co-founder to do something. I just never clicked with somebody to the level that we could start something. So it’s very much that versus it’s not like any calculative thing.

It’s just, I never got to it. You know, if it happens in the future, it’s great. I love building businesses. I mean, don’t get me wrong. I love, I get so much joy out of it. So let’s see how things go. I mean, I’m happy at Arize. I think it’s a great company. I think I found home here. So I do want to, you know, build this business and, you know, hopefully help make another iconic company.

Siddhartha Ahluwalia 39:47
And can you tell us some process that you built at Cloudflare for building a go-to-market team? Like what are the parts of the go-to-market teams that you built, you know, a journey from 10 to 50 specifically, and this is helpful for founders who are listening and also from who are going from 1 to 50 million, they have PMS.

Manish Jindal 40:08
Right.
So in terms of like how the go-to-market team was at Cloudflare, right? Yeah. So I mean, I think our go-to-market team was fairly standard in the sense that the different roles we had, like, so we had the traditional SDR or BDR role, which is their job is to, because we would get a lot of inbounds.

So their job was to work on the inbounds, qualify them, schedule the meeting, pass it to the AE. Then we have account executives. Account executive, initially, we really hired like a more mid-market kind of account.

Siddhartha Ahluwalia 40:42
Always in Bay Area, SDR and AE.

Manish Jindal 40:44
Yeah. So Cloudflare, we were very big on FaceTime. Like, you know, until the COVID time, we had multiple, we have multiple office, but everybody was supposed to come to office five times a week, like Monday through Friday, it was required.

So we had, initially we had office in San Francisco. And then the second office we opened was in London, basically. But initial go-to-market team was all at San Francisco based.

So we had SDRs, we had account executive, we have solution engineers. These are basically your kind of the pre-sales people. Then we have a couple of people doing partnerships.

Then we have like a very traditional post-sales team, which is customer success manager and customer success engineers. And then we had a customer support team for the L1, L2 tickets, basically. So it was a traditional model from a team structure perspective.

But the profile of the people we picked was more unique in the sense that we did not just hire people based on experience. We hire people based on like the hustle and people with their owner’s mindset and people who kind of have the chip on their shoulder, they really wanted to do something. We hired a lot of people out of universities.

We hired a lot of the SDRs out of Santa Clara University, and they did tremendously well for us. We hire a lot of mid-market AEs. We did not hire your traditional enterprise AEs early on.

Siddhartha Ahluwalia 42:14
Why?

Manish Jindal 42:16
I think that the reason for that was that we wanted to build a fast velocity motion, which was like the deal size was about $50,000. And we wanted to do deals like with the sales cycle less than 30, 40, 50 days, basically fast motion. And our worry was that if you bring in a lot more enterprise people, they may push the company in an enterprise segment faster than we want to go, basically. And so our intention was to bring people who are going to drive a lot of the fast velocity go-to-market motion, and we’ll bring in the enterprise people when we think our product is ready or the company is ready for it. Because at the end of the day, the kind of people you hire has a big influence on where the company goes from there. So in terms of even hiring the leadership team in the go-to-market, we were hiring people who were more open about doing things in a different way versus coming with a playbook.

So we never hired somebody who was like, I have a playbook. This is how I do things. We were like, no, we don’t need you because we were like, this business is unique that we want people who are like first principle thinker who just look at things and then make the call of what is the right way to go about it versus like, hey, I need to build a segmentation of this SMB, there’s a mid-market, there’s enterprise. For enterprise, I need to put people on the ground around the world. Then mid-market, I’m going to put here. We’re like, no, none of that. We don’t want that. So we were very intentional about that.

Siddhartha Ahluwalia 43:41
And one more thing I want to check with you. Now, I’m hearing constantly with founders that GTM in the AI world is changing. But what GTM has changed or team structure has changed in Arize as compared to Cloudflare?

Manish Jindal 43:53
Right.
The GTM, I think, you know, how the GTM is different in the AI world. I mean, I think if you’re selling AI, then the way buyer go about is very different. So you can look at it in two different ways.

One is how the GTM is different for a traditional software company. And I’ll come back to that in a second. But if you’re selling, if you’re an AI product, right, what’s the difference is that in a traditional older world, you always have to go displace an incumbent.

There’s somebody spending money on something, you are actually saying that, you know, you move that spend from that vendor to us. So it’s very much about you kind of are a disruptor and you’re telling the world that, hey, we are faster, better, cheaper than what you’re using, come to us. Right.

But in the AI world is you are not displacing spend. You’re not displacing vendor, actually. All you are doing is telling them that we have built something which will allow you to either make your AI work investment or AI is going to may help you in either reduce your cost or drive the top line. Right. So means that the traditional motion of like, you know, very consultative selling and whatnot doesn’t apply as much to the AI product of the past. What really important now is the education, right? Because everybody is hungry for education around like, hey, how do I go build agents? What are the right tools? Where do I get started? How do I make sure my system doesn’t hallucinate? How do I scale it? You know, like those questions are there.

So means that your go to market, you have to lead with education more and education, which means that you spend a lot more time in building documentation. And also your brand has to be more educational. And then you can attract the developers to start using your product. And if they start using the product, then you use those developers as a way to go wall to wall with those customers. Right. So how you lean in is slightly different. So, you know, the role of DevRel becomes way more important than what was like.

Siddhartha Ahluwalia 45:54
And you are saying because of AI, the budget didn’t existing, for example, in case of Arize, Eval and AI observability never existed. Yeah. So we are not displacing any budget, actually.

Manish Jindal 46:05
We are actually, in a way, are part of this bucket of the AI investment they are making. We just want a sliver of that. So we don’t have to go say that, hey, go remove Datadog.

No, like keep Datadog. You know, but if you want to really care about scaling your agents in production, you need the right tooling. And we are one of the one of the tool for that. Right. So it’s a different conversation, for example. So that way, the go to market is slightly different. But then if you’re selling to enterprise, right, you still have to go to the same procurement process, the security.

Siddhartha Ahluwalia 46:41
But are you selling to mid-market or enterprise?

Manish Jindal 46:43
We sell to enterprise. So we, our strength is large enterprises.

Siddhartha Ahluwalia 46:47
And why, you know, your specialty initially, what you built at Cloudflare, going deep in mid-market and then go to enterprise. Why did you choose enterprise where enterprise, as you said, they bend the product.

Manish Jindal 47:03
Right, right. So, you know, why it happened here is, so we have an open source product, Phoenix. And what we realize, very interesting inside, actually, I learned this after coming to Arize, is that the biggest, like if you think about, you know, you would think that every developer would want to use open source.

But what we realize is that it’s the developer at the larger companies that wants to use open source. And if you are a developer at a, let’s say, born on the web company, you actually want more of a off the shelf product. It doesn’t have to be open source. And the reason for that is if I am, let’s say, work for a bank, I’m a developer. I don’t, I care about my data. So I don’t want to move my data to the cloud. So open source is easy. I can use that. Right.

So for us, what happened was because of open source, we got a lot of developers from very large companies start using us. And in a way, then that just pulled us into those accounts. So as an example, Wells Fargo is a big customer of Arize.

And they were before Wells Fargo became a large customer of Arize, they were like 100 developers using the Phoenix product. And that’s the reason why we were able to get into the Wells Fargo. So in this case, we went to an enterprise not by choice, but because customer pulled us into that direction. Right. That’s kind of, you know, one thing happened. And the other thing happened also for us was, so we started as a observability company.

And then now we are a more end to end solution where we do both evals and observability. But observability, you need it when you have something in production, your agent is in production. And enterprises like the large enterprises where the agents are in production more than, you know, like if you look at the Silicon Valley companies, they all are building agents, but very few of those agents are in production actually.

But if I am, let’s say bookings.com, if I’m Uber, if I’m DoorDash, I actually have agents in production. So I really need a real observability solution. And that’s one of the also reason is like those customers pulled us into the mix. So that’s kind of how the journey happened for us. But like from our perspective, I think we are very, very strong with large enterprises. But now because we have a very strong evals product as well, we are now winning in the digital natives as well, very much.

So now at this point, actually we are doing really well across a lot of customer segments, which is a lot. So we as a company are, you know, trying to like keep up with the pace of like, you know, like we are getting pulled in all the directions. But the good news is that we have a product for like a lot of different kind of customers right now.

Siddhartha Ahluwalia 49:35
How big is the team right now?

Manish Jindal 49:36
The overall sales go to market team is a little over 60 people right now.

Siddhartha Ahluwalia 49:41
60?

Manish Jindal 49:41
60.
Yeah.

Siddhartha Ahluwalia 49:42
And overall team?

Manish Jindal 49:44
Overall Arize about 150 people.

Siddhartha Ahluwalia 49:46
The large part of the team is go to market?

Manish Jindal 49:48
Yeah, go to market team is quite big now. And it’s growing really fast.

Siddhartha Ahluwalia 49:51
And everyone is based here in SF?

Manish Jindal 49:53
No, it’s a distributed team. So we have, like I think the critical mass is in the Bay Area.

Siddhartha Ahluwalia 50:00
For the go to market team?

Manish Jindal 50:01
Go to market, a lot of in the Bay Area, but we have people around the US as well. Then we have a team in Germany, we have a team in UK, we have a team in Singapore. So we are now truly global go to market.

Siddhartha Ahluwalia 50:17
But isn’t it too early to build a global go to market?

Manish Jindal 50:21
Right, right. I think in our case, what happens is the customers are pulling into that direction, right? So for us, for example, like we have a lot of customers out of UK now, and there’s just massive demand. Similarly, even before we went to Singapore, we now have a lot of big customers in Asia, and they pulled us in.

So it’s more that instead of us intentionally going and winning the market, it was the customer pulled us into those markets, basically. I’m a big believer in when you think about the go to market is that you never want to go into a market cold. Because what happens when you go into the market cold is that you have to invest a lot of amount of money to warm that from a marketing perspective, you have to hire the sales team and whatnot.

And it takes them a while to get the first customer. So if you think about, you know, you added all this cost into the model, right, marketing costs, and the people cost and whatnot, but the revenue comes like six, nine months or 12 months from there. And if you if you for whatever reason, you pick the wrong market, now you wasted time and you wasted a lot of money, right? So so if you can find like attraction in the market, for whatever reason, in our case, we found traction in like like Asian market and also UK and German market, that it was a no brainer for us that we already have customers who love us, we already have a lot of open source developers who love us. Like it’s a no brainer for us to put people in the market. So it was a no brainer decision for us. But if that wasn’t the case, we wouldn’t have gone into those markets.

Siddhartha Ahluwalia 51:44
And the folks that you are hiring, you know, to lead, do you have a similar function SDR and AEs?

Manish Jindal 51:51
Right? Yeah, we have, we call them BDRs. But yes, same BDRs and AES.

Siddhartha Ahluwalia 51:54
And what kind of profiles are you looking when you’re hiring them at AES?

Manish Jindal 51:58
The BDRs?

Siddhartha Ahluwalia 51:59
Both.

Manish Jindal 52:00
So I mean, BDRs, I think BDR profile varies, but we are hiring a lot of out of the college people as well for BDRs or with a one to two years of selling experience. For AES, we are hiring some more experienced AES because you know, we sell into the large customers. And typically, AES will have experienced five to 10 years of selling experience.

Siddhartha Ahluwalia 52:21
In the enterprise space?

Manish Jindal 52:23
Yeah.

Siddhartha Ahluwalia 52:23
And do these AES need to come from selling plumbing products or you’re hiring application AES also?

Manish Jindal 52:29
Yeah, we kind of all, we want to hire AES who are hungry, who hustle and who knows how to run a sales process, who know how to do multi-threading and all, you know, who know how to do sales. What they have sold is less important, because we can teach them how to sell what our product, we can teach them our industry. So we are not like, only hiring AES who actually are coming from a similar space, we have AEs who come from very different industries.

Siddhartha Ahluwalia 52:55
And let’s say for a exec today looking to join startup, either in infra or in application space, right? Because there’s so much noise in AI world, like the principles that you laid out earlier in our conversations, extreme PMF, right? Plumbing, great founders, how do they figure out these things about a startup, you know, when they are exploring too?

Manish Jindal 53:22
Right, like, yeah, I mean, I think there’s no magic formula, you have to just spend a lot of time with the founders.

Siddhartha Ahluwalia 53:30
Like you spent at Insight Partners.

Manish Jindal 53:32
Right, yeah.
So you have to go spend a lot of time with the founders and then pick the ones where you actually click and feel that, yeah, this is the right home for me. I don’t think there’s any shortcut per se. But I mean, some of the shortcuts, I mean, if you really, you basically, you know, if you want to make sure that you have the right founder, you ask the question during the interview process, right?

And you sometimes ask questions where it might be irritating to them, and then you see how they react to it. Right? And then you get a sense for like, you know, how well they take the bad news, how well they take the criticism, you know, like, you know, those things you do want to assess.

So there’s few things you can do in the process. But I don’t think there is any shortcut to just spending time with the founders and see where you are able to build the comfort, right? So never, like, rush into a job, I feel, and you take your time.

Siddhartha Ahluwalia 54:28
Now you are creating the real plumbing for AI agents that Arize. So there’s a narrative that a lot of AI is hype, only 95% of AI agents are, you know, getting into production, they start feeling. What’s yours? Because now you sit at the back.

Manish Jindal 54:46
Yeah. I mean, AI is not a hype. Like, I mean, I think we all, I mean, we should, we agree on that. AI is not a hype, but the statement that a lot of investment went into AI and the results are not what we expect them to be at this point. I mean, there’s little to show for it. And part of the reason is that they just, the companies have not put the right tooling in place. You know, like a lot of companies made this mistake of just using the existing tool to make AI work, which did not work. But now I think you will see a lot more agents going into production and doing really, really well because now, you know, like we see with our customers, you know, where their volume is just, is growing exponentially right now. So I think we are reached to that inflection point where we should start seeing agents being used a lot more effectively than what is in the past because the right tooling is in place.

You know, you definitely, you know, like if you think about how can you have an agent running at scale if you don’t have the evals in place, like otherwise you are totally flying blind. And especially when the system is non-deterministic, which can do, you know, you have no idea whether it’s performing the way it should be. And that’s part of the reason why a lot of these agents failed in production because the right tooling was not there.

So now I feel that the companies are putting the right tooling so we should definitely see a big, big increase in the usage of agents.

Siddhartha Ahluwalia 56:17
So what, you know, to get AI agent to work in production really well, apart from eval and observability, what kind of other tooling do you require?

Manish Jindal 56:26
I mean, I think, I mean, if you think about broadly, if you think about the agents, what do you have is a few things. You have your LLMs, you have your tools, then you have your like rack system, right? So you basically, and then you have the orchestration where you build the agent, right?

So these are the things you have to get it right, basically, like the piece around when you’re building your agent, like let’s say you use orchestration framework to build the agent, you have to put the right evaluation system in place. So to make sure that your agent is doing what he’s supposed to do. So as an example, right? Agents are, AI systems are non-deterministic, right? Which means that for input, it will give different output every time. So now how do you know that your agent is doing what he’s supposed to do? So the way you figure that out is you actually quantify the output of your agent into different metrics, which, you know, and then you test your agent to make sure that it’s, those metrics are true to what it’s supposed to be. So the metrics could be less hallucination. You want to make sure that it’s not hallucinating. You want to make sure that there’s a correctness of answer. You want to make sure that there is no PI information. There could be like hundreds or thousands of different metrics you need.

So first you need to have the right system in place where you actually can run those evaluation at scale. That’s the first point. And now let’s say you are happy with it. Now you move that agent into production. So now in the production, now another thing gets introduced where now in a traditional software, you actually are restricting the user to interact with the software in a certain way. You know, you restrict like what kind of inputs you can give to the software.

But in an agent, the user can interact in a natural language. Now you introduce this point, like how the prompt being asked could be very different than what you tested in development. So now whatever you test in development, that needs to be continuously tested in production. So those evals becomes the online evals. So it means that every time that agent gets called, those evals runs to make sure, yes, agent is doing what he’s supposed to do. So that’s where the concept of observability came about, which is, you know, in a traditional observability, which is for a deterministic system, what it’s good at is telling you whether your system is doing what it’s supposed to do.

You know, the yes and no, right? Whether is your service up or not, right? Whether let’s say latency is high or latency is low. So it’s kind of answering yes or no. In the world of agents, if you think about it, you could be in a world where you actually, agent is perfectly, you have a data dashboard and everything is green on the dashboard. It’s doing all the right thing, but the agent may be giving you totally garbage hallucinated answers.

And the system will say, oh yeah, it’s all green because the system is running perfectly fine. It’s all green light, everything. So now in production, you actually need a observability where it can actually tell you whether, not that the traces are right, but where the problem may be. You know, so let’s give you an example. If you look at the agent, agent is a multi-step of workflows, right? So, you know, you make different tool calls. You are retrieving context from let’s say your RAG. You actually are calling other agents as well, right? So the answer is not that you actually were able to make a call. The key thing is, did you make the right call or not? You know, did the context you got from RAG was the right context or not, right? Did you actually make the right tool call or not? So it’s less about, did you make the tool call? It’s more, did you make the right tool call? And that’s where, you know, the AI observability come into place.

It tells you that actually you made the wrong tool call or the context you got from your RAG retrieval, actually that context was wrong or your context came right, but the answer you got was totally different than what the context was fed, right? So those things are so important to make sure that your agents are performing as supposed to in production. And that’s where, you know, this comes about is like, you know, you do a question, what do you need in production?

You need all the systems in place. Then only you can be sure that your agent is going to do what he’s supposed to do. And if it’s not, then you can go fix it very, very quickly.

Siddhartha Ahluwalia 1:00:56
And let’s say for enterprises today, what use cases they qualify, you know, because now AI agents are hype and people will build AI agents left right and center, compromising their security, people are also giving their teams to spin up AI agents. You know, they create a central platform, for example, you know, like Glean, and companies will give their teams power to spin up AI agents. So how do they realize that whether they need an AI agent for a specific use case or not?

Manish Jindal 1:01:28
So your question is like, where do you put the agent? I mean, I think agents is like nothing but workflows. I mean, I think where I mean, you want to put agents in any anything where the steps are kind of defined to do the job, right. So I’ll give you an example. So one of our customer is DoorDash. And DoorDash has one agent for refunds.

So let’s say, you know, you get a DoorDash order delivered to you, and you’re not happy with the order or something wrong was delivered to you. So then you can ask for a refund. And now, if somebody asks for a refund, there are certain steps you take, like, you know, even if you’re a human on the back, certain steps you take to make sure whether this is right or not, and then you issue the refund. Agent is a perfect thing which you can do, right. So DoorDash is using Arize actually for that different agent where agent actually asks for, can you take a picture of the order? So, for example, you take the picture of the order to really say what is like, is the right thing was delivered or not, then it actually goes and check the history of that particular person in the past, like how often do they ask for refunds or not. So based on that, actually, it can predict with a lot of certainty whether to give a refund or not. Right. So that’s a perfect example where now why you need to human for that agent can easily do it. So those are examples where, you know, it’s there’s information needs to be processed by a system or a human. You use that information to go check on something. And based on that, you make the call. Agent is perfect for that. You know, so those kind of scenarios, I mean, we see working really well.

Siddhartha Ahluwalia 1:03:03
And these agents are for DoorDash are created by Arize?

Manish Jindal 1:03:06
So agents are created by, so we are the layer for evals and observability. So agents do create it by, you know, they can, you can create agents through like frameworks like let’s say Langchain or CrewAI or what, or you can actually, you know, just write your own engine. But we are the evals and observability layer for those agents to make sure that, because if you think about, right, if agents, if DoorDash agents start to perform, like start hallucinating and let’s say if they start accepting all the refunds or if they start rejecting all the refunds, it’s a big problem for DoorDash.

So they actually have to make sure that agent is doing what it’s supposed to do. And it’s accurate, you know, so that’s why observability becomes so important. Like if you can’t, you can’t just allow yourself to be in a black box like that. And we have seen so many examples, right, in Air Canada, which is a customer of Arize. Before they were Arize customer, they launched an agent to help with booking the tickets. And it was mainly around booking the tickets through miles.

And people will find a way to like give it prompts where actually the agent end up issuing miles to those customers and then those customers use those miles to buy start buying tickets. And Air Canada had no idea that’s happening, you know, because they didn’t have the right tooling in place. And eventually they realized, so they did shut down the agent, but then they got sued by those customers and then they had to actually honor all those customers with those tickets.

So then they came to us and then now we put the right, you know, the evals and observability in place. And now that agent is doing really well, right? So the idea is like, you know, if you really want your agent to scale in production, you need to have the right tools in place. And first you test that in development, then you have the right tool in place, then only the agent is going to, you can be sure that it will deliver the business outcome you want it to deliver.

Siddhartha Ahluwalia 1:04:56
So it’s a great point that you made, but as Arize, how do you discover customers like Air Canada who have a specific pain point? Because it’s a great pain point, right, to solve. And then if you solve for such kind of a large customer, then the customer would be willing to pay at scale millions of dollars.

Manish Jindal 1:05:18
Right, right, right. So I mean, I think either customer discover us or we go find the customer, it’s a typical, right? So for us, we are very fortunate that we have an open source product where that is driving a lot of awareness in the market about Arize. And so a lot of the customers find us through, for example, open source or the word of mouth. But then a lot of customers do search, right? So we invest a lot of money in AEO and SEO and making sure that a customer can find us.

We invest a lot on education, we invest like education through DevRel, that helps. So there are like a lot of things, there’s no magic wand per se, you have to do a lot of things where you become discoverable by the customer. Other thing we are doing is we also have been very strong partnerships with AWS, with GCP, with NVIDIA, with IBM, where they also introduce us to a lot of their customers.

Because if I am, let’s say GCP, I really want the consumption to go up for my LLM and the consumption is only going to go up if you have the right system in place for agents to really perform. So it’s in their best interest to bring Arize into the mix. So it’s like typical go to market, you have to find a way to build distribution and awareness in the market. And there are a lot of ways you can do it.

Siddhartha Ahluwalia 1:06:38
But this is amazing. Like what you are showing is that there are real problems on the infra side and how they’re affecting customers. Like in this example, in case of Doordash, agents are hallucinating and issuing refunds in case of Air Canada, where action of an agent can get them sued.

Manish Jindal 1:06:58
Yeah, or a big financial implication or reputational risk, right? So that’s why I know the narrative in the market is that AI is not doing what it’s supposed to do. But I think it’s not that. It’s more that the people just have not taken the time to put the right tools and tooling in place.

Siddhartha Ahluwalia 1:07:16
So when do companies like Air Canada decide that for a certain workflow, let’s say booking tickets through miles, can they hand over from a human to an AI agent when they are 100% sure?

Manish Jindal 1:07:29
I mean, I think, you know, so what I’ve noticed is that human in the loop is still a thing, right? I think, you know, we still don’t have the trust for agents to be fully autonomous yet. So which means that if, let’s say, today it takes you 100 agents, let’s say support example, to process your support tickets, maybe with the right agents in place, AI agents, you need 10 people and 90 other people can be replaced by the agents.

But you still need some human in the loop to make sure it’s doing what it’s supposed to. And over time, the trust gets built and everything, you know, you work through all the corner cases and then maybe it can become anonymous. But even until today, it’s not. I’ve not seen any agent to be like 100% autonomous yet. I think we will get there. It’s like autonomous cars, right?

I think in 2015, like 95% of the technology of autonomous cars was figured out. But that the last mile, it took like 10 more years to get to the point where now actually they are autonomous. I think agents will be the same way, it will be faster than autonomous cars, but it will take another few years for it to be like, really be the truly autonomous agents.

Siddhartha Ahluwalia 1:08:41
In your opinion, then the current, you know, we discussed about it. The system of record layer is not going anyway, but the current hype where SaaS companies like HubSpot are down by 80%, you know, ServiceNow is down by 50% whereas, and it happened because a private company, Anthropic, released a new version. So is it like at one area, the narrative is the agents are not delivering for 95% of demos are failing. And the other end, the narrative is because Claude can build everything now. That’s why, you know, the companies that are leaders in their category are dying.

Manish Jindal 1:09:24
Right, right, right.

Siddhartha Ahluwalia 1:09:24
It’s very confusing.

Manish Jindal 1:09:26
It’s confusing. But you know, if you ask me, they are the two agents in the world right now, which are like truly the most powerful is the claude code and the cursor. Those agents are amazingly, insanely good.

So I think it’s a matter of time that agents which are built by companies start doing what you expect them to do. As long as they have the right tool in place, I think we will get there. I think in the next 12 to 18 months, we’ll see a massive inflection point.

Now, we are as a company are betting on that, right, where we have so many customers who are using Arize, and we expect that the consumption of those customers should go like 5 to 10x in the next 12 to 18 months, right? So like, if you take example, Sierra.ai, right, Sierra.ai got 200 million ARR last quarter, and then they added 50 million ARR the next quarter. And it’s not that they actually signed $50 million new deals.

What happened was the agents they deployed in production for all these retailers, and now these agents start to see a lot of volume. And they are charging based on deflection, right? So more deflection, more the money they make.

So I’m pretty sure that this $50 million ARR they got the next quarter, a big chunk of that is actually existing customers just start using more. Same thing I think we should happen with Arize, where as the agents start to go into production, we should also see a big jump in the consumption. But that’s what we are betting on.

And we are already seeing those signals. I mean, it’s very clear if you see from the Sierra of the world or Decagon of the world it’s happening.

Siddhartha Ahluwalia 1:10:59
And now, let’s step into your heart as an angel. You would have received like in the last couple of years, maybe a few hundreds or thousands of pitches from founders. How did you choose the 15 companies that you choose to invest in?

Manish Jindal 1:11:14
Yeah, so for me, how I chose, I chose the company where I understood the space. I think, you know, like example wise, like Atomicworks, for example, right? I invested because I understand ITSM. I understand that ITSM is a space where there’s a bunch of steps needs to be done every time and agents can easily do it. Right? So I understand the space and I understand the application and I see the pain.

And then I know the founders really well. So I just felt that this feels right. So for me, it’s like if I really understand the space and understand what is agents trying to do and I can wrap my head around, then I’ll go invest in it. If I cannot, then I just won’t basically. Right? I mean, if you come and say that, oh, this agent can do the XYZ in health care, I don’t understand health care. So I don’t know. Right. I remember somebody pitched me this idea of this agent can read the x-rays and in a typical way, when the x-ray happens, the doctor only look at the area where you have the pain and just check that and it doesn’t look at the rest of the x-ray.

So it’s just, you know, there’s a lot of information is wasted away. And the idea was that this agent can read the whole x-ray and create a database. And actually then it can extrapolate, you know, like based on all this information about the health of the person and whatnot. And I’m like, sounds like a good idea, but I just cannot wrap my head around how this is going to work. So I will not invest in something like this. I’m sure it’s a great idea, but I would not. So for me, something tangible that I can understand, like what it is.

Siddhartha Ahluwalia 1:12:44
What made you invest in Composio, for example?

Manish Jindal 1:12:46
Yeah.

Siddhartha Ahluwalia 1:12:47
Or Portkey.

Manish Jindal 1:12:48
Yeah. So Portkey, because the middleware. Like, I mean, I’m a big believer in middleware. I was like, you guys are going to sit in the middle of the application and the LLM. I think I’m going to invest in it. And I felt that Rohit and all, just like smart guys, good guys. And I, you know, that’s kind of why I went with Composio. Composio, why I invested, I think Composio changed their ideas, actually, when I invested to what they are now.

Siddhartha Ahluwalia 1:13:13
What was the idea when they, when you invested?

Manish Jindal 1:13:15
Composio. Gosh. What was Composio’s initial idea?

I mean, Composio’s initial idea was more like, you know, like how they are building different APIs to be able to bring like data in one place, something of that sort. But for me, Composio, I invested because of the founders. I just like those guys, you know, they are from IIT Bombay and like humble and super smart, you know.

And I felt that, you know, it was early enough that the valuation was low. I was like, I think I can bet some money on these guys. It was like very much investing on the founders than the idea.

But surprisingly, of all my investments, that’s one of my better investments. So at the end of the day, I think, you know, the question becomes, are you betting on the horse or the jockey, right? So in this case, I invested in the jockey and it turned out to be a good investment.

Siddhartha Ahluwalia 1:14:05
Yeah, because the idea changed.

Manish Jindal 1:14:06
Idea changed. Yeah. Same thing with Cloudflare. I think if Cloudflare also came through a business school competition and the idea was very different at the business school level. And now what it became.

Siddhartha Ahluwalia 1:14:17
What was the business school?

Manish Jindal 1:14:18
The business school idea was very much about like a more like ad revenue. So you basically build the network and a lot of traffic and then you can make money through the eyeballs. That’s kind of the idea. Versus like actually to be the security, the network layer. So it was very different monetization idea versus what it became over the years.

Siddhartha Ahluwalia 1:14:38
And when you joined at employee number 45, what was the product and the idea that they were selling?

Manish Jindal 1:14:45
So at that time, it was three things, CDN, so content delivery network and DDoS, which is distributed service and DNS. And the idea was like setting as one bundle and because nobody was doing that.

Siddhartha Ahluwalia 1:14:59
I think they are still the leader in this..

Manish Jindal 1:15:01
Still the leader in that.
But that was that was what at that time was was the then the WAF came about and then all the things came about. But initial idea was very much like let’s have the the best bundle accompanying and lie in the market and at a very, very reasonable price.

Siddhartha Ahluwalia 1:15:19
In your opinion, in Neon portfolio, we have seen the highest amount of acquisitions, profitable acquisitions happen in the infrastructure space. I’ll share one example, a company called Requestly that marks HTTPS requests. They got bought out by a browser stack, a 5 billion dollar unicorn in India.

We have a company called Zenduty, which is an SRE platform that competes with PagerDuty that we entered at 6 million valuation. They got bought at 25 million valuation by a PE called Xurrent, a PE bank company. Similarly, a company called Logiq, it’s a Bay Area company in observability, competing with Splunk on some observability use cases. They got bought by a PE called Apica, a PE bank company. So why do you think, I’m still trying to figure out, we have five acquisitions in Neon portfolio. Is it incidental with us or is it?

Manish Jindal 1:16:20
Right. I mean, acquisition is the real thing right now. I do feel that any company right now, any company you pick, the likelihood for them to stay independent in the long run is very, very low.

Siddhartha Ahluwalia 1:16:38
Why do you say that?

Manish Jindal 1:16:40
Yeah, because I think the reason is the market is changing so fast and a lot of the things you are building kind of end up becoming a feature in a bigger stack and you’re better off becoming part of the bigger stack versus stay as a feature because you can become irrelevant. And it depends on the founders, right? What is the aspiration? But a lot of the founders feel that, hey, the market is moving fast. What I’ve built is working today.

What if it doesn’t work in one year? Because the AI is moving so fast, there’s a lot of things which make sense today may not make sense, right? But if you’re getting acquired, the other company can get a lot more out of it and you can get a decent outcome. So people are doing that. So I get that. But for me, for example, when I’m an investor, I’m okay with that.

But when I’m picking a company to work, I want to pick a company where their aspiration is not to get acquired, but as to stay independent. Like Splunk and Cloudflare, I want to stay independent. I think Arize is a company I do see founders, because one of the founders is a second time founder. He has taken a previous company public as well. So for him, it’s more about really making this company work and building something tangible. So I’m excited about that. But you never know, you know, like for the right price, everything is sellable, right?

Siddhartha Ahluwalia 1:18:00
Yeah.

Manish Jindal 1:18:01
But that’s happening for sure. I mean, you hear it all day long and at a very massive elevations that’s getting acquired. One of our competitor, Langfuse, this company out of Germany, I think they had like 13 employees and they recently got bought by Clickhouse. And I mean, they did not disclose the number, but we believe like it’s close to a billion dollar acquisition.

Siddhartha Ahluwalia 1:18:23
Wow. 13 people.

Manish Jindal 1:18:24
13 people. Yeah.

Siddhartha Ahluwalia 1:18:26
Any revenue?

Manish Jindal 1:18:27
They had a revenue. I think they were like 6-7 million revenue.

Siddhartha Ahluwalia 1:18:29
But they were not a good, they didn’t get acquired for their revenue, obviously.

Manish Jindal 1:18:33
No, I think they got acquired for the technology. I mean, but yes, but they got bought for like almost a billion dollars.

Siddhartha Ahluwalia 1:18:40
But today, the other narrative is that any technology is not a mode. Any technology…

Manish Jindal 1:18:46
Technology is not a mode. You’re right. Yeah. Yeah.

Siddhartha Ahluwalia 1:18:48
So then why are larger companies buying smaller companies when technology is not a mode anymore?

Manish Jindal 1:18:55
I think because they can build something larger, like, you know, they can build a basket, you know, like one particular thing may not be a mode, but the basket is a basket, like, you know, where we had competition from, like, different CDN providers within the DNS providers and DDoS provider and whatnot. And but if you go ahead to, like, if somebody just wanted DDoS, we may not be the best solution. Like, we were a great solution to DDoS, actually, I take it back.

But like, you know, for some pieces, we may not be the, have all the bells and whistles vibe, you know, other company, but we have the best basket. And I think a lot of these companies who are acquiring, they feel that I want to build the best basket out there, then I’m not displaceable, basically. But within the basket, few things may become obsolete or become like commodity, and that’s okay.

But the basket will still have a lot of value. I think that’s what’s going on.

Siddhartha Ahluwalia 1:19:42
Thank you so much, Manish. I really loved our conversation, learned a lot. We’d love to do part two sometime, you know, of this conversation, where I can dive deeper into some aspects.

But completely, you know, thoroughly enjoyed the conversation.

Manish Jindal 1:19:56
Thank you. No, it was a great conversation. I enjoyed it as well.

Yeah. Very thoughtful questions.

Siddhartha Ahluwalia 1:19:59
Thank you.

Manish Jindal 1:20:00
Awesome.