Afternoon General Session | MongoDB.local London 2026

It's a good first question. From my standpoint, you know, yes, there is a lot of talk about compute being commoditized, storage being commoditized. There is this thing about large language models being commoditized, SLMs being commoditized. And if you truly think about artificial intelligence, the intelligence has to come from data. So I absolutely believe that data layer will not be commoditized. Actually what I would argue is that data layer has been the only constant since the tech industry started where you start with the mainframe 60 years ago, 70 years ago with IBM and many other companies and they had different databases. Then you had the Oracles and others that came through. So while internet transformation, cloud transformation, and now AI, data layer has been the only constant. And it's still a large market that's growing and it's not commoditized.

So I started my career at a data company or database company. It's called Oracle Corporation. And I understand somewhat how it works and why people build applications or whatever they build on top of a database. And then when MongoDB originally reached out saying, 'Hey, would you consider being CEO at MongoDB?' And I said, 'Wow, that's interesting.' It's like full circle starting from Oracle as the first job out of university and coming here. And what I was surprised is that not only frontier model companies that I found out later on who use MongoDB, but like you mentioned Eleven Labs, they were here this morning, great, they use MongoDB. And of course large banks here in the UK, many of them for mission-critical workloads including payments use MongoDB. So that was a positive surprise for me to understand that I thought they would bucket us in this NoSQL category. Yeah, there is SQL, Oracle and NoSQL, but it is used as mission-critical workloads. And now when I think about AI, I feel that there are just more conviction after speaking to many, many customers, startups, and others is that there are three things that will be constant in this AI world that I have now learned. Of course LLMs are here to stay, whether it's LLM, SLMs, you know, people talk about different type of models, open-source models, whatever it is. Then you have the data layer and data layer is here to stay. And agent layer, whatever you want to call, wherever you want to draw the boundary, that's here to stay. So that is a new learning for me. Everything else, whether you look at application software, other types of software is in question.

For me, unless you have completely transformed yourself. So ChatGPT somewhere in 2022 fall and then all the advances in Claude over last few months including Codex and others. From my standpoint, these companies that you talk about, unless you start with AI first principles and figure out are you going to leverage AI to create a brand new product because just to infuse an AI and say I'm an AI company now, I'm rebranding myself as an AI company, that does not work. So either you use AI to create brand new products that are relevant in this world or you really, really make on AI first principles a completely different company and reinvent yourself. That is the only path to growth and unless you can accelerate growth like Eleven Labs was 350 million in December and Mattie posted 500 yesterday or day before yesterday in 4 months, unless you show that acceleration of growth, these companies will have a very hard time.

First I want to say one of the frontier model companies that I just spoke to last week. Okay, they are very well known. Cannot use the name but you can guess. But when I spoke to them last week, they said, 'CJ, we are behind in selling to enterprises.' And I said why are you behind? And they said we thought that our software will just sell itself and we don't need a quota carrying seller and it'll just sell because we are AI, that's the message we are trying to give, and we found that was very, very hard. So we are accelerating hiring people because still today it is bought by people at least in the B2B context, it's bought by people. So at least in 2026 if a frontier model company is saying that, Harry, I personally think that tells you a lot today. In the future if agents are buying, assuming they get to the right level of deterministic behavior and so on, then yes, 100% you want to get ready for any company that is trying to sell to a business what would it mean in that world for an agent to buy you, what would be the valuation criteria because today it's still Excel with weightage across every line item pricing. I'll go to Gartner or industry analyst and get a pricing that is this fair market value and so on. That whole thing needs to translate in the agentic world for agent to then just make a decision. Because I'll tell you even on database today people will say, 'Oh, I prompted my favorite tool and said which database should I use and you know particular AI tools said you should use this database, okay.' And then it's a bank and the bank is like, 'No, no, no, no, that's okay, that ChatGPT may have told you this but we have only these standards because we run mission-critical applications here.' So even if agent is recommending this particular database or this particular software tool you cannot use it because this is the control and governance function we have. So those are the multiple levels that you have to pass through before an agent can just make an autonomous decision to buy something.

Yeah, I would say the more regulated the industry, the higher the bar, right? So if you're in a regulated industry, the banks, healthcare and so on, you have to follow a certain set of rules because you are in front of regulators all the time, right? Sometimes once a month with multiple countries that you do business in. So for those regulated industries today, they have data sovereignty requirement, they have this requirement, that requirement, so on. You can't just show up and say, 'Hey, use this AI product that automates your legal department or finance department or HR department.' They are going to say, 'Hey man, we really need to understand all this really well.' So the first thing is the higher the regulation, the higher the adoption.

100%.

I would say where the stakes are lower like internal employee facing tools or productivity, that place they are adopting knowledge worker companions and so on. But true client facing, so say you want to make a determination on mortgage and you want to say I will approve, the agent will decide whether you get a loan or not and then we have to prove to the regulators the audit and other things, that's where the adoption barriers are much higher.

Some, some, because we speak to enterprise customers. MongoDB has many, many enterprise customers. So we speak to them all the time and it depends. There have been firms whether it's in financial services or healthcare or public sector who are on the bleeding edge and they say, 'Hey, we need help here is where we are, here are the use cases we have identified that needs to be either AI enabled or we need to create agents.' And CJ does your team have forward deployed engineers? Like even Eleven Labs was saying that they have FDE teams that go. Brett Taylor does the same thing with Sierra as you know because the skill set on the other side is not there and it is all about the ROI. Meaning Brett, and I'm sure he'll be okay if I say this, last year when he was pitching Sierra to me he said, 'CJ, I'll get you to the first outcome in less than 30 days and I'm going to put my top four forward deployed engineers so that we help you build these agents rather than you building yourself so you get to the support outcomes that you're looking for.'

I have two ways, Harry. I've been entire career selling to enterprises, the products whichever products we had. And on one hand is agent just another application? And if you think the old model you had system integrators who help you build application, then you had Indian system integrators who would customize and maintain those applications for you. So you look at those and now with AI, is forward deployed engineers basically the same thing that these people are going to come help you set up once but they also then need to maintain. So on one hand I want to say ADLC or SDLC what people call software development life cycle or application development life cycle where consultants, contractors, engineers help you build, design, build and maintain. If that same thing happens in AI because it's just another piece of software, I want to believe that yes they will be there for a long time to stay.

Well Palantir is the greatest joy is like if you're a CEO and you need to be AI first you need to drop 20 to 30 million bucks for it to mean anything at any scale.

Can I ask how much of Mongo's code today is AI versus human?

Yeah.

20 to 30% tops. So that 20 to 30% that window has shrunk significantly now with all this coding assistance. So then you say okay can I take that to the bank? So if you were a shareholder or investor, I know you're an investor, but if you ask me that question I would say if every single engineer if I average out or median 30% they are more productive that means two things. Either I increase the innovation velocity by 30% but keep the same engineers. I keep the innovation velocity the same but I may not need this 30% of engineers. Or as the company continues to grow and our chief revenue officer wants to sell more we want to create new products, these folks will write brand new products using AI first principles. That's how you look at it and that's how I look at it in terms of what does it mean to be 20%, 30% more productive with this coding assistant.

I want to be thoughtful. My entire company...

Yeah. We just took out minutes. I would say we have a lot of imagination in the company on other areas of the data platform that we can build ourselves and to build we still need human engineers and I want to give them a canvas on which they can paint as in write code and create great products. And if AI tools are helping them that's great but I would say we would have actually more engineers, not as many as I thought we would need 3 years from now because it's a growth company. We are growing every year. We have more and more customers that we need to support. They ask for requirements, features, this and that. So we will have more engineers but not exponentially or linearly. Our revenue growth, whatever it is over next three years, our engineering headcount will not grow linearly with it because of AI.

I would say I thought about this question actually. So there will be no ums and uhs.

Yes, I did think about that one. I would say at the highest level, MongoDB, you know, the most modern database got created just 19 years ago and there are a lot of areas that are adjacent to where we play and we want to create those products in those areas because we believe we have great teams that can now leverage AI. So if I have truly unlimited budget, first is innovation. The product platform has to be world-class and as these enterprises, because whether you ask Mark Benioff or somebody else, the TAM is really selling to the businesses for companies like us. There are a lot of things that we could disrupt in the market that we can build ourselves with AI. So this is the one thing I would do.

That I cannot share but you should just think, you should just think about as some of our competitors try to get in our space we want to get in their space and we feel that we have the unfair advantage to do so.

Yes. So I am going to pass that.

Yes. Pass. Good pass. But second thing I would say is also you asked about the go to market question. We are not ready today for agents to buy software on their own and there are geographies that we would like to invest in that we have not invested as the company grew with public market pressures and so on. So that's what I would do from unlimited budget perspective.

I would say even if you look at the software industry's history since 1970, so now it's 50 plus years. That was the easiest model. Seats was the easiest model. Hey, it's Microsoft Office. How many employees you have? They didn't even have Office back then. It's Microsoft Excel. How many employees you have and who is going to use Excel. And you look at today and you look at next 15, 20 years between cloud and AI. I would say the value that these enterprises are asking, they do not want to buy because one of the recent, if there is a recent history, lot of people bought lot of things that they did not need during the pandemic era, right? So you see many, many software companies say, 'Hey, it was overbought so our retention rates are not high, our churn is higher.' That would have not happened if it was not seat-based pricing model. So I am a massive fan of consumption-based pricing model. That is the right way to go. And even I look at the new age AI companies they are all saying the same thing that either it's outcome-based pricing or consumption-based pricing but it cannot be seat-based pricing because you really do not know what would be the future of that company. You asked me, 'Hey, will you have more engineers?' The truth is I don't know how many more. I know I will have more but I don't know how many more. These companies want to sign a three-year contract, five-year contracts based on seats. As a CEO, I would never sign it.

You know, I'll start with something just recently and of course I have recency bias. Just kidding. You see the announcement from Anthropic today, right? This morning or last night, whenever it was in the US.

I forgot this was recorded.

No, no, no, no. This is going good so far. I think I'm doing okay.

You look at specifically that specific partnership that got announced, it's all about the capacity. Where can I get the capacity? The new data center that's not even being built, we are going to use that because we see that demand for next 3, 5 years whatever it is or 10 years. So on one hand when I look at that, that is not what I expected a year ago that this is how much people will talk about neo clouds moving to on-prem. You know, I had a customer, Fortune 50 customer, they are one of the largest spenders on this particular hyperscaler, do not ask me which one, one of the largest spender on that particular hyperscaler and they said they are now moving their workloads to on-prem, the data center they were going to decom because that particular hyperscaler, the CEO from that hyperscaler said we cannot give you more capacity in the regions you are asking for and too bad, like literally. And they said, 'Wow, we spent this much money, it's in B as in billions, we spent this much money with this hyperscaler and the CEO is still saying I just, physics, whatever you want to call it, I just don't have that.' And so they are doing that. That I did not expect a year ago that this is what the world will come to in terms of data center in space or...

I mean you have it on both sides. On the customer side yes that's what they're worried about and there was this whole 10 years ago approximately, 'Oh, data centers are dead, 100% of the applications are going to move to public cloud.' And then it drove the growth in the hyperscalers. And today the way I look at it from both sides, if you're a software company what does that really mean from your architecture standpoint, if you're 100% in hyperscalers, what does that really mean? If the large language model, the three, four frontier model companies, they become the next hyperscalers for 10, 15 years, that's the signs they are exhibiting right now. What does that mean for you as a software vendor? But on the other side, if you're a customer, a Fortune 50, Fortune 10 bank right here in the UK, you really have to figure out what does leveraging AI mean and will you be stuck like the example I just shared with you someday where you're locked in by a hyperscaler or by a frontier model company and you don't know where to go.

Yes.

Oh. Lots of choices. I would say because I'm originally from Bombay. It was called Bombay when I left. So now it's called Mumbai or something. The IPL team Mumbai Indians.

We would sponsor MongoDB because one I'm from there. Just kidding. Second, we have a massive developer community in India. Everybody watches IPL and that would be great.

Oh wow.

You know Harry this is live right?

Yes I know. I know.

Yes. I would say because and I know you have discussed this with many other guests in the past. The speed at which things are moving right now are so fast. It like the cloud era, the internet era that happened over last 30 years, that's nothing from my standpoint. The compounding change that's happening is so fast and somebody on the board who is current, who is technologist and who can keep up with all these things and provide some kind of guidance to me and the management team. That would be super helpful to do that so that they can say okay yes you are reading as much as I'm reading, we follow everything that's happening whether it's frontier model companies, AI native companies or just tech and how it's changing in the industry. That's, I would say we would benefit from that.

What did I change my mind the most on last 12 months? I would say I definitely originally when I joined even MongoDB I said okay we are the core database company, people run their... you know one of these customers in Spain told me that, this was in Barcelona, 'CJ, one-third of payments go through MongoDB and if something goes wrong in your software you will have many, many folks in Spain very unhappy with you.' Okay so I mean that's a high bar, that's the accountability and privilege we have as an infrastructure software company. And I said okay we should be always very systematic in terms of how we release the product, it should be secure, durable, all those things. And now I'm like if AI is writing X% of the code which will go up over time, right now it's sub 10 but it'll go up over time, how we still... AI being table stakes for the engineering team is like I would say the biggest pivot I made because we are learning a lot from customers who are trying to also move fast and they say here are the agents that we are building and this, so we need to be fast even though we are at the infrastructure software level. Like not being fast is just not an option because most infrastructure software company leaders will tell you that you have to be very systematic on how you move forward because you can bring down companies, nations or governments for the use cases they run infra software.

I would say I have a very simple parallel from my standpoint whether you look at industrial revolution or whichever revolution you want to look at, the speed is so high. And in next 5 to 10 years, so MongoDB, human developers loved us, right? And they still love us and that's why everybody's here. But I think it will be orders of magnitude more agents that we don't even know. Over long term, 5 to 10 years, I think AGI becomes real so there is lot of work that gets done. And coming back to your first question, I think data layer will always be the constant in the next 5 to 10 years and that's the opportunity we have and that's what gets me really excited that everybody should use our data layer and the AI world will truly deliver on its promise. So I'm an optimist there.

That's fantastic. Thank you very much.