Snowflake CEO Ramaswamy Predicts ‘Zero Chance’ AI Agent Era SIs Look The Same

“I see this as an opportunity for us to do better with the most progressive of the GSI and RSI partners,” Snowflake CEO Sridhar Ramaswamy said.

Snowflake CEO Sridhar Ramaswamy said he and his team have active conversations with partners on the future of their business model in the agentic artificial intelligence era–with smaller partners sometimes showing the ability to change faster than larger ones.

Comparing global system integrators (GSIs) and regional system integrators (RSIs), Ramaswamy told a CRN reporter during the AI and data cloud vendor’s Snowflake Summit 2025 conference that the shorter-term engagements RSIs favor has allowed them to adopt AI tools faster in some cases than GSIs working on longer engagements.

Snowflake’s service partners “know they have to change” to survive in the agentic AI era, Ramaswamy (pictured) said during a press conference. “We have active collaborations with our partners. … I see this as an opportunity for us to do better with the most progressive of the GSI and RSI partners.”

[RELATED: Snowflake Summit 2025: The Biggest News In AI, Agents]

Snowflake Summit 2025

Summit 2025 runs this week through Thursday in San Francisco. CRN has collected Ramaswamy’s boldest statements across his keynote speech and Summit-related press conferences, wherein the CEO–who’s held his position for about 15 months–weighed in on channel partners, AI agents, his company’s pending acquisition of Crunchy Data and more.

Erik Duffield, CEO of New York-based Snowflake partner and IBM Consulting subsidiary Hakkoda, told CRN in an interview that agentic AI, migration work and data governance projects continue to drive customer IT spending despite macroeconomic headwinds.

In health care, data sharing has facilitated continuity of care across providers, for example. In supply chain, AI is modernizing suppliers and vendors to the end of the product life cycle.

AI systems at customer sites show “the interchange of data, the value creation of when that data meets your employees or your customers,” Duffield said.

“That includes monetization use cases, but also experience, net new revenue, cost savings,” he said. “It's no longer just transformation within a company. It's those companies banding together to modernize how an entire industry works.”

Ramaswamy said that his team’s work with SIs has included leveraging forward-deployed engineers, employees that have a mix of product and field systems experience. The CEO has preached creating a migration factory engine within SI businesses.

“All of the GSIs and all of the management consulting firms that these GSIs often have as arms understand that there is zero chance that things are going to work the way they are working today in five years,” he said.

The following are Ramaswamy’s comments and insights on Snowflake’s direction and strategy, the AI and data analysis technology markets, the impact of AI on the channel, and more, taken from his keynote speech and statements made during a press conference.

AI Is Changing SIs

The interesting thing about these [services] partners is that they know they have to change.

Certainly, if you look at the folks that had big federal contracts, for example, let's say it's been a learning experience for them over the past few months in terms of how to operate.

We have very active conversations with them about what does the future of system integration look like? And what are things that we could be doing together to shape that future together?

The challenge for them, as is the challenge for us with our professional services team, is how do you drive that kind of disruption fast enough?

We have a set of techniques. We are actively experimenting with things like forward-deployed engineers, which are this interesting combination of both product people, but one with field systems experience.

We are investing in a pretty heavy way into what does a migration factory of the future look like?

Migrations have mostly been defined by problems that engineers thought they wanted to solve, and much less by what problems do you actually need to solve to get a migration?

No one likes to do testing, so tools for testing data [that has been] migrated are not that great. But if you think about it, this is mission-critical data, it better be the case that the new system behaves like the old system.

We have active collaborations with our partners. … I see this as an opportunity for us to do better with the most progressive of the GSI (global system integrator) and RSI (regional system integrator) partners.

I would say the RSIs … are better placed to navigate this simply because they are less about super long-term engagement and setting strategy. And It's less about a lot of people on the ground doing things with a very distant horizon.

And they tend to be much more about–how do I get this project done in six weeks? And if you give them a better tool and tell them you should do this in three weeks, they're just as happy to do it in three weeks.

There's some nuance in how the different folks operate. But I think all of the GSIs and all of the management consulting firms that these GSIs often have as arms understand that there is zero chance that things are going to work the way they are working today in five years.

Snowflake In The AI Era

[The AI Era] feels real.

Even something as simple as my ChatGPT account being linked up to Apple Intelligence–awkward as it is to invoke it–is still a little bit of magic.

A lot of what Snowflake is doing is making this magic happen for enterprises, for the data that matters to them. Snowflake has already built both a product and the reputation as being the most trustworthy analytics platform that there is.

The best enterprises in the world store their most valuable information with us. And we're helping them unlock value from that, both in terms of simply getting it to business users faster without going through BI [business intelligence] tools or through data analysts. But also, now [they are] beginning to think at a higher level about what should the future of workflows be?

My conversation has gone from, ‘I can reduce your Teradata costs’ to, ‘We can be a strong ally in how you rethink how your business should operate.”

You will see a lot of us practicing what we preach in terms of AI getting embedded into how our systems operate, whether it is with making migrations go just a whole lot faster, or even things like ‘how do you make notebook development faster using modern copilot-like experiences?’

We are continuing to invest in strengthening the core data platform on which all of this sits. We acquired a company called Datavolo roughly this time last year. It's going to get rebranded as Openflow.

And the cool thing about that is, all of a sudden, many data sources–over 100 in fact, both structured data sources but unstructured ones like SharePoint and (Google) Drive and Box–all of the data can now be brought either into cloud storage or into Snowflake and then be part of workflows that can then be created with search and indices and models on top of that data.

Snowflake’s AI Pricing Model

We have a consumption-based pricing model where we charge for units of compute that are consumed. And then when it comes to AI, like other other folks, we charge by token.

It is very value aligned. It does not waste resources from our customers, as in, we don't sell a set of licenses that might or might not get used.

Most SaaS (software-as-a-service) licenses are estimated once a year. And most of the time there's a lot of wastage that comes with it. And I can understand why some folks will say that they want outcome-based pricing on higher prices, but we think the consumption-based pricing model is one that aligns value created between us and our customers.

For some kinds of products, we might consider other models. But with respect to data and AI and the products we have created so far, we are very comfortable with the consumption-pricing model.

We don't have a large subscription business to protect. And that's part of what gives us the freedom to be able to go innovate and, perhaps, disrupt how things should work.

AI Unlocking Unstructured Data

Many companies have given up on their unstructured data repositories. They look at every contract that they have signed for the past 20 years and go, ‘There's not much I can do with it.’

The power of AI models in general to unlock that data is pretty profound. … You still need quality signals about which ones you should look at. Maybe you have signed 20 versions of contracts with the same company over the years, and so you need semantics for how you disambiguate between them. But I think that unlock is pretty big.

There's a lot of data that is sitting in legacy systems, and practical difficulties that people have with these systems include … it's a box. It has some data. It's a good system. But you came up with an idea to do something else, or you want to juxtapose that data with some other data that you have in the enterprise.

With Snowflake, all data that belongs to you as an enterprise is implicitly joined. You have to figure out semantics, but it's one system.

That's not the way that it works with on-prem systems. But on the other hand, migrations out of these systems have been long and painful. We've started pretty ambitious efforts to automate some of this migration because AI and agentic flows–in the spirit of practicing what we preach–can be very, very helpful.

You will see us lean a lot into–how do we make data migrations go faster?

Migration is a fancy way of saying you're going to work for a lot of time and be at the same spot you're in a year from now. But on the other hand, now the exec staff, the CEOs, understand that if you can actually bring data into a modern system, you make it AI ready.

Part of our message is data that is in Snowflake is AI-ready data. That's a trend that you will see us lean into, that legacy migrations will accelerate because people understand now that it can be a tool to transform business, not just do better reporting.

Snowflake Third-Party Partnering

We have customers that use Snowflake directly, but when it comes to applications, we have relied on third parties to build applications on Snowflake.

We don't pretend that we are AWS [Amazon Web Services] when it comes to being an application platform. We let our partners build what are called native applications, which are specialized.

Agentic AI is going to blur the difference between what integration work that used to be done by system integrators. … They made data flow from this one system to this other system, and then sometimes they created visual interfaces that you and I would call applications.

What is remarkable about agentic AI is the promise that those should be a whole lot easier to create, and things like custom UIs [user interfaces].

Their ability to build high quality UI with just English instructions is remarkable. What you're going to see is a big blurring of the lines between what an application is, what an integration layer is, and what used to be a data platform.

Our strength comes from the purity of our mission. We are centered around data. We don't really sell anything else. We don't sell GPUs [graphics processing units].

Ours is a much simpler framework. ... We want to let agent components be created on Snowflake. If you have an important dataset and you want to make it interactive and expose that as what's called an MCP [Anthropic’s Model Context Protocol open standard] component … We are perfectly happy with that.

If there are two or three data sources whose data has been put into Snowflake and you want to create a workflow on the data, we're going to let you do that. We want to play this very much with our strengths and with an eye toward interoperability.

There's going to be both competition and also cooperation in these areas. I met with Swami [Sivasubramanian, AWS’ vice president of agentic AI]. … His top topic was, ‘How do we make sure that we don't balkanize agents so that these agents cannot talk with those agents?’

The industry will cooperate to make sure that there is that kind of interoperability. And then … the best products are going to win. The people that make the best models today are not the ones with the most money. There are hyperscalers who have more money, and yet, there are other people that are the best at making models.

We think of a high-quality data platform in a similar light. We are comfortable with our position, and also very careful to not pick unnecessary fights. I see partnerships as a win-win. When it comes to agents, we will absolutely stress interoperability.

Keeping Up With Rate Of Change

For a while, people were convinced about ‘scaling laws.’ They said, ‘If you just train a bigger model and put in more compute, you will get better.’

That clearly hit a ceiling sometime last year. But the attention, very much, has shifted to post-training.

I don't think this space of innovation is slowing down anytime soon.

It is hard to keep up. But on the other hand, I think there is more certainty about what's the value that you can get from AI.

The days of us using [the ‘find’ shortcut keys] Command [plus] F to figure out what some PDF is saying about a specific topic–that's long gone. You just don't have to do that anymore.

Whether it's unstructured data, whether it is structured data, whether it's specific kinds of image generation or some kinds of video generation, voice recognition, the frontier has gone way past.

I talk a lot with both boards, with CDOs [chief data officers], with CEOs about how should you think about AI innovation?

You have to experiment a lot. It is mentally tiring. But on the other hand, I think just keeping abreast of the things that are eminently doable is also … a good thing.

It's very hard for startups that are building on top of these models, because all of a sudden there's a model upgrade, and it has gone past what they have done. I genuinely think that that's a very big challenge.

Enterprises are at a better spot, and Snowflake is in a better spot because we think of AI as an accelerant on top of the data that people are already bringing into Snowflake anyway.

We are not in the business of experimenting with foundation models and what they can do. It lets us be pretty effective.

Customer Security

Data security is a joint responsibility. We are a platform, but we work with customers to keep their data safe.

We have introduced things like making it easy for administrators to say that all Snowflake users, for example, need two-factor authentication. Doing that is easily the best thing that administrators can do to protect data.

We're going to make it mandatory for every account to have two-factor authentication. We've also started doing things like supporting more modern methods for two FA, like face ID. I don't know if you folks use it, but because it relies on physical presence as opposed to even an application or a dongle, it's considered somewhat more secure.

We are also surfacing data and alerts to our customers in terms of how many accounts they have that are not protected in this fashion.

Because we are a single, unified platform, all of the rules that you have—whether it is for data access or data masking—works out of the box with the AI products that you create on top. And then with AI products, obviously, customers have to worry about things like malicious people trying to get chat applications to say improper things. And so we offer things like Cortex Guard to protect against malicious input.

When it comes to customer data itself, we offer a flat guarantee to our customers, which is customer data is customer data. We never use that data for training any model ourselves.

All of our customers rest assured that their data and any AI products that they build on top of it will only be used to answer questions for them. Snowflake will not be benefiting from that in any way.

ROI-Positive AI

With any feature that we launch, we focus on customer ROI. What we have learned over the past few years, especially from the excess of spend in the pandemic era, is that we need to be very measured about how we deliver value for our customers.

The very first thing that I tell our customers about Snowflake and AI is you should only launch ROI-positive products. You need to make sure that you're getting value for your dollars.

We are using AI to make our enablement material much easier for our sales folks. We are creating tools with Snowflake Intelligence that, again, makes it much easier for sales folks to get the latest information about their customers, whether it is spend on Snowflake, updates on use cases that they have gotten from partners, transcripts from a meeting that a colleague had. All of that information is now accessible from within Snowflake.

And we'll also obviously support interoperability. So if they need to read data from someone else's agent or some other data source, we absolutely will facilitate that.

All of these set us up on the right track for how do we drive change within enterprises? It will start with the obvious—things like, you can create a pipeline faster. Or you can write a piece of SQL query faster. It'll move over to chatbots, which now thousands of our customers have deployed.

And then you get into compound systems where the agent can decide which data source that it should consult. And then you go from there to, ‘can you now be able to also do updates to other systems from right within Snowflake?’

And then things like low-code applications that you can spin up right within Snowflake Intelligence ad hoc to get some tasks done.

It starts with having valuable data, then giving access to the data faster, then driving insights on top of the data and then streamlining workflows that are a part of every person's job. The line between SaaS software and what an agentic platform can do is going to blur over time. And we feel very well positioned to take advantage of this arc of change that honestly is rushing through the enterprise ecosystem.

Keep AI Simple

The true magic of a great technology is taking something that's very complicated and making it feel easy. The key to a great solution is simplicity.

You can choose to check boxes on functionality, string together solutions and pass along all of the complexity to other people, the end user. But actually, complexity creates risks. Complexity creates cost. Complexity creates friction and makes it harder to get the job done, whereas simplicity – it drives results. And that's why we want simplicity at the heart of how we design products.

We make it easy to share, collaborate, view and build with data. We make it easy for you to get value from your data, whether you're a data analyst or a data scientist or a business leader.

And that simplicity has never been more important than it is right now in the age of AI.

You should be able to ask a question with a voice memo and get an answer on your enterprise data. You should be able to launch a customer app without having to write the line of code. You should be able to harness the world's best AI models to create agents that are tailored for your business.

Thousands of you are using Snowflake AI and machine learning to accelerate things from insurance underwriting to analysis of investment reports. Underpinning all of these results is a strong, capable data foundation.

There is no AI strategy without a data strategy. Data is the fuel for AI. And Snowflake’s AI Data Cloud is powered by a connected ecosystem of data. We have thousands of customers sharing data, apps and models from one another on our Snowflake Marketplace. We have over 3,000 listings from over 750 partners.

You need data you can trust. That's why Snowflake has placed such a huge emphasis on delivering quality and accuracy in AI. It's why Cortex has a market-leading, 90-plus percent accuracy when it comes to AI chatbots. We are not stopping here. We know you need to move ahead with confidence that the right people are using the right data for the right products.

With Snowflake, we help you do more with your data end to end. From the inception of data to getting insights from it, we help you at each stage of the data life cycle.

Crunchy Data Acquisition

Snowflake has agreed to acquire Crunchy Data—a leading provider of open-source Postgres technology—to help us deliver, build, innovate and scale using the tool.

Postgres is truly amazing, and it's going to be right inside of Snowflake so that you can get the performance, the governance and scale that Snowflake is known for with the familiarity and openness that comes with Postgres.

We are doubling down on that momentum behind Postgres and investing in its future while preserving the openness, the extensibility.