Products Get Replaced, Platforms Stick: MongoDB's CEO on the SaaS Bear Thesis

Sarah Guo opens by asking the question every investor asks: what is software worth when you can generate a lot of it? CJ Desai calls it a spicy opener and answers by zooming out to earlier transitions, from mainframes to the internet to SaaS, which he notes has been around for at least 25 years.

Some companies say their moat is a great customer relationship or channel, but Desai's own answer is speed. When a technology transition arrives, he asks, are you building as fast as you can and learning from the shift? He points to Meta's pivot toward mobile in the early 2010s as the kind of fast move he means.

Companies that fall behind will keep being asked what their future is, by investors and by customers alike. Not every bet will work, he says, but assuming the terminal value of some software is zero is, in his view, overblown.

Asked why concepts like stickiness feel abstract to engineers, Desai answers directly: platforms are sticky and products are not. He recalls his old hiring manager at ServiceNow, Frank Slootman, saying "tools are for fools," and says any product can be replaced in a disruptive software market. Positioning as a platform may lengthen the sales cycle, but it makes the purchase a considered decision.

Guo pushes back that startup advice usually favors a narrow wedge, such as the IT service desk at ServiceNow. Desai agrees you need a killer entry use case, but argues that an entry point that was easy to get into is easy to leave when nothing has been built around it. Growth from zero to 100 million is one thing; getting from 100 million to a billion and then to 10 billion is much harder.

He says the target is Fortune 500 enterprises, where banks and insurers have operated for 100 years or more, and where multiple products, security checks, governance and integrations pile up. As an example, he tells of a London bank CTO who said 300 applications ran on MongoDB, only for Desai to learn the denominator was 9,000. The more they use, the more embedded MongoDB becomes in their infrastructure.

Guo raises the common worry that some of a bank's 9,000 applications will be generated on demand by vibe coding, tailored to each company instead of bought as standard software. Desai concedes that app velocity goes up, but says you still need a go-to-market channel to reach a bank with its big technology budget.

Banks, he says, talk to regulators more than to customers and vendors, so they ask whether an application will pass regulatory tests. They also want multi-cloud resiliency rather than something built only on AWS, and sometimes a truly sandboxed, air-gapped on-premises setup. Healthcare and public sector buyers bring similar demands.

Asked what incumbents like ServiceNow or Cloudflare should do, Desai says the first test is whether your total addressable market still looks large. Then, he says, understand your moat and use AI to strengthen it: integrate with more systems, build products for new use cases faster, and show re-acceleration of growth. Innovating more without selling more, he warns, signals trouble.

That is what investors look for, he adds. Unless a company shows re-acceleration, they will say they are neutral, or in extreme cases that they are bearish.

Guo notes investor dollars have shifted from business software to AI infrastructure, model companies and hyperscalers, while data and developer infrastructure got less attention. She asks how Desai, who had options and was at Cloudflare, picked his next move. His first criterion is a durable TAM, informed by starting his career at Oracle and learning how a database platform, middleware and applications scale.

At MongoDB, he says his diligence with many customers surprised him: the database is behind mission-critical apps such as retail e-commerce, commercial banking, healthcare and insurance claims. Databases have been around for 50 or 60 years, while MongoDB, created in 2007, is about 18 years old and, he says, the one truly disruptive force in that market.

He also heard from digital natives of the 2010 to 2015 era and from AI natives that they build on MongoDB. He argues that the data in AI applications is unstructured, messy and needs high velocity and search, something the company's founders almost accidentally fit. The cloud transition is still going on, he says, and the AI transition has only just started. He quotes the previous CEO, Dave, saying no other next-generation database has passed a billion in revenue, let alone two billion.

Guo summarizes the bear case: investors worry about SaaS applications, about data infrastructure because of how quickly applications are built, and about AI-native companies because value might end up in the models. Desai agrees that this is the dominant thesis. He says that since late 2022 he has never seen the future of software questioned like this, by investors and by customers deciding between vendors.

His answer is to name what will remain. LLMs will be part of the stack for the foreseeable future, as will the agentic software framework, and a data layer is needed to store data. Everything around those layers will evolve, and it must show real value, for instance an AI-native company that deeply understands insurance use cases and shows fast speed to value.

Desai says he talks to at least 10 customers a week and treats a week with fewer as a failure. His read on large enterprises is that adoption is still slow. Office-productivity copilots drew lukewarm feedback on value, while coding assistants, which took off in 2024, produced very positive feedback, and he calls 2025 a breakthrough year. Customer support is still being tested, and he says end-to-end customer experience isn't fully handled yet by AI-native companies.

He relays the question customers ask him: should an AI-native vendor be an "and" or an "or" next to a system of record like Salesforce? Desai says he asks vendors whether they are a layer on top or a full replacement, and that a promise to replace the incumbent for less, faster and better, with disruptive pricing, gets his attention. Guo calls that surprising, since conventional wisdom says systems of record are too embedded.

Within MongoDB, he says its CIO, Deepa, is approached by AI-native companies many times a day, and he wants the company to be AI-first, transforming the business rather than being merely more productive. He also cites a European retailer he met at NRF in New York that, after failed ERP implementations, decided to build its own system on MongoDB.

Guo asks when Desai, a product and engineering person, began thinking so much about business strategy. He credits John Thompson, then CEO of Symantec, around 2005, whose bar for how to interact with, sell to and serve customers was very high. The advice was to ask customers about their other problems and pain points, in person if possible, because it lets you see around the corner and pattern-match.

He gives a recent example: a retailer CTO at NRF said e-commerce was 20% of revenue and was happy with MongoDB, but did not know the company also offered search and vector search. Desai says that kind of intimacy teaches how customers deploy a product, how long it takes and how much value they expect. He says product people who think "build it and they will come" miss this, and he prefers to name the person at the bank who made the bet, not just the bank.

On surviving technology transitions, such as MongoDB's Atlas cloud shift and now AI, Desai says comfort is the danger. At ServiceNow he told skeptical engineers that not leaning into AI was not an option. He compares it to Nokia and BlackBerry, which sold well even after the iPhone launched, and calls transitions mostly a change-management problem.

Guo adds that incumbents often bundle products and relabel a piece as AI or cloud, so leaders need guard rails for intellectual honesty between what 10 customers say each week and what the market sees. Desai says that when MongoDB reported its third-quarter results he was asked repeatedly if they were due to AI, and he said no: it was the core. AI-native companies are building on MongoDB, but few have reached scale, and he sees AI as additive to a core business that is still growing.