AI is turning every company into a software company

As AI capabilities accelerate, nearly anyone can create software. The real business constraint is no longer code, but judgment: how quickly an organization can learn, adapt, and redeploy talent at scale. In this episode of The McKinsey Podcast, Snowflake CEO Sridhar Ramaswamy joins McKinsey’s North America Chair Eric Kutcher to dig into how AI is industrializing intelligence, supercharging productivity, and forcing leaders to rethink everything from pricing models to workforce strategy.

In this recurring series on The McKinsey Podcast, Eric speaks with top CEOs about the practice of leadership.

The McKinsey Podcast is regularly cohosted by Lucia Rahilly and Roberta Fusaro.

The following transcript has been edited for clarity and length. To watch the full-length version of this interview, visit The McKinsey Podcast playlist on McKinsey’s YouTube channel.

When anyone can code

Eric Kutcher: You are at the center of this moment with AI. In your own words, how do you see the world today?

Sridhar Ramaswamy: AI is going to have a fundamentally profound impact on the cost of software creation, which, in turn, means that it’s going to have a profound impact on how information is used. It will affect how information is processed, what it means, what intelligence means, and what decision-making means.

It’s a set of changes cascading out from the very central impact on software and information. The only historical analogies that come anywhere close to this are things like the printing press or the internet, when it comes to content.

The special aspect of AI is that it also brings intelligence. I call it the industrialization of intelligence, along with these models. Packing all of that together leads to a concentration of change that we are trying to get our heads around.

Eric Kutcher: You’ve always been an engineer. Over the past six months, how has engineering changed from a software point of view?

Sridhar Ramaswamy: It’s a little hard to predict what the outcome is going to be, so I’ll make lots of predictions to compensate. One prediction is that many more people will be able to write software. I already have people on my sales teams, for example, who don’t think twice about shipping a new visualization dashboard or a new application, because why shouldn’t they? All they have to do is describe what they want done in English and out comes the application. I also have geniuses on my team who, as far as I can tell, are 50 to 100 times more productive than they were even a year ago. This is leading to the rise of a class of “uber programmers” who are enormously more capable and who intuitively understand what it means to use the power of coding agents to create software.

We also think about coding concepts, not just to write code but also to develop agents that can critique code and improve it—for example, to address software security issues. That level of sophistication is still evolving.

This is having a profound impact on how software will be created. The inevitable is that software as an industry is going to be transformed massively. It’s no longer going to be the cottage industry where people can trust that a certain population has adopted a piece of software and is therefore not going to switch. I think the ability to create new things is just so much greater. I expect a lot of innovation, but also a lot of disruption.

Eric Kutcher: If output increases 50-fold, what does that mean for product cycles, pricing, and customer education?

Sridhar Ramaswamy: All of those are difficult problems. A positive aspect of AI is its conversational nature. As long as you conceptually know what it is that you want to do, it’s quite easy to learn by just asking the coding agents. The natural interface for software is changing from bespoke web interfaces, in which apps made by individuals are stylized and designed to serve the masses, to more fluid, conversational interfaces.

That is a huge option. At this point, you should treat most tasks the same way you treat traffic. First, you ask Google Maps what you should do, and then you decide whether to take that route or another. I think that’s something all of us still have to internalize.

Eric Kutcher: This is also the most democratized technology we’ve seen. People learn by interacting with AI—asking, refining, debating—and improve outcomes in the process.

Sridhar Ramaswamy: That’s the magic of the moment. The process of creation is very interactive. You assign the AI agents different personas. You can assign different personas to agents, have them critique each other, and refine outputs before sharing them with humans.

I wrote a thesis about the agentic enterprise that we have since turned into a product called SnowWork. The first version emerged from a series of conversations. It was effectively one long conversation with our coding agent that I turned into a document and then shared with a number of people. Then, I pitched that document to several CEOs and took their feedback and input it back into the original to write a new version of the document. After that, I created a deck using a coding agent based on the updated document. Just the amount of AI that goes into creating a document about AI was kind of surreal and unbelievable.

Rewiring the economics of software

Eric Kutcher: We are also living in this world where software companies’ valuations have been under pressure. Most would argue it’s because the terminal value, which made up 90 to 95 percent of the value of the whole enterprise, has been downwardly adjusted. How should we think about the value of software today?

Sridhar Ramaswamy: What will emerge over the next few months is insight into the structure of what will and won’t succeed. For example, I spoke about us being a consumption model. This means we don’t have to price things ahead of time.

Pricing is always one of the most difficult topics on the planet, because all kinds of things can go wrong. The consumption model is inherently beneficial because customers only pay if they get value from what they’re receiving.

There are also a number of questions open about what stays beneath the model versus what lives above it. In other words, yes, they can create software. But will disruption happen in software that is closer to what end users interact with, perhaps in enterprise or in consumer software, or will it exist in a lower layer?

Eric Kutcher: The consumption-oriented model, by the way, is what the major LLM [large language model] providers are using. It’s not a totally new model. We’ve seen other companies—there were some in cyber for quite some time—focused on data ingestion, which you could argue is another form of consumption.

It’ll get interesting because you are starting to run into companies that are saying, “Wait a minute. This thing is growing at a rate and pace that I don’t fully understand, and it’s not predictable.” Someone else also pointed out to me that not every token has the same value.

Sridhar Ramaswamy: Correct.

Eric Kutcher: My personal view is that the pricing model we’re starting with is unlikely to last long. There will be pushback that will put real pressure on whether you can do it as a pure tokenization.

Sridhar Ramaswamy: All of those things are natural consequences. Some products will go the way of subscriptions. But the important thing to remember right now is that coding agents, especially the ones that are taking good programmers and “50-folding” them, are incredibly powerful because the demand for those candidates is very, very elastic.

With products like Snowflake Intelligence, we are experimenting with consumption-based pricing, but also with having a per-person limit to prevent unexpected runaway spending by individuals and a per-account limit.

These are the kinds of innovations that will come. In fact, some folks who code in Cursor have deliverers that charge both a per-user fee and a consumption or token fee. To your point, I think there’s going to be a lot of innovation.

And, yes, there will absolutely be pushback if ROI does not keep pace with cost. What’s unique about the moment is that so much value is being created, but there’s pressure on core AI models, including frontier lab models.

Open-source models are getting better. And while we work with the big model makers, we also host open-source models. If they’re as good as their big model counterparts, our customers will want to use them instead of using frontier models. We can confidently say it is very early, and you really do need to predict quite often.

Eric Kutcher: Sridhar, you talked about using AI in the context of software development. In your CEO circles, what are the other transformative things in other parts of a business that get you excited, particularly as we think about growth as an avenue toward productivity?

Sridhar Ramaswamy: Two surefire hits of AI with massive ROI: One is software engineering. This is where Cortex Code [CoCo] sitting next to Snowflake is a big deal, and it will have a big impact on all of our customers.

The other area is support. There’s a lesson there. We basically wrote our own support system. The team wrote the code in six weeks on top of Cortex Code. We launched with very little fanfare, and it had a massive productivity improvement. Our support queues are essentially empty because we can crack through those problems.

Our SRE [site reliability engineering] team rebuilt the alerting function for their observability platform on top of Snowflake and Cortex Code. Again, there was little fanfare and a huge impact. Problems that used to take them four days, crawling through Kubernetes logs—which is the worst punishment you can give an engineer—are all automated now.

Rethinking roles, not just workflows

Eric Kutcher: This feels more like a business transformation than a technology transformation. How are you approaching change management?

Sridhar Ramaswamy: First and foremost, we demystify the technology. Because we do this, we are the beneficiaries of a really lucky experiment we ran by accident. We created Cortex Code, our coding agent, but we didn’t restrict the coding agent.

We have a command-line version, which some people don’t like. We also have a desktop version, which lots of people like. Snowflake also happens to have a lot of data in it that’s available to every employee. That’s long been a tradition in the company. We say, “Hey, whatever data you need to store about Snowflake, you put it into Snowflake.” That internal tool is called Snowhouse. That made the entire company AI-literate very quickly, in a way that I did not have to mandate.

I didn’t have training programs. CoCo was so popular that it spread virally because people got so much utility from it. Beyond getting the core creation and selling motion right, it was about getting your team to embrace AI without forcing AI down everyone’s throats.

That was a happy accident. Now we are in a phase of figuring out what AI means for each function. For example, if we no longer need a demo-making team because every executive can make their own demo, we move the demo team into other roles within Snowflake.

There is an element of figuring out what things you need. What are the new jobs that are being created as a result of AI? How do you help people transition over to those jobs? That’s the process that we are going through.

Eric Kutcher: A question I receive a lot is, “How is this world going to evolve? Are we going to have smaller organizations?” I find it interesting that you just pointed to two very real examples where you didn’t say, “I’m going to have fewer developers.” You said, “I’m going to develop more code, and therefore I will grow my value proposition to my clients in an outpaced way. I have real opportunities for people I haven’t been able to fill before, because I had to consider tasks that are no longer necessary. Now I have an opportunity to put those candidates in places where I can drive more growth.”

Sridhar Ramaswamy: That’s absolutely the way I want things to work. That’s the first option. The other side, and I won’t pretend that this is not happening, is that there are people, even at Snowflake, who are struggling to adapt. They are used to doing things a certain way.

We try to be nice about it. As a leader, I will try every technique possible to motivate people. Clearly, it’s going to make a difference in things like their performance management. But I also appeal to a more fundamental instinct.

I’m a software engineer by training. My sons, 26 and 24, are also software engineers. I tell our technical people, “I was and am terrified about their future employment.” I push them very hard to make sure they’re at the cutting edge of how development should be done today. A lot of my appeal to our own Snowflake team is not only to make Snowflake better but also to protect their livelihoods. This is a time of massive change, and you need to embrace it to thrive in this world.

Lessons from a varied career

Eric Kutcher: You’ve helped build start-ups that went on to be acquired—including at least one very successful exit—so you’ve seen what it takes to grow a company from an idea to a business that someone else wants to buy. You’ve also worked inside one of the largest organizations in the world and have led teams at a massive scale.

How do you compare and contrast those two experiences? What did you carry forward from each that shapes how you lead as a CEO today?

Sridhar Ramaswamy: They’re very different experiences. Sometimes you only learn by doing. Being at Google for 16 years was an amazing professional experience. I joined as an engineer when Google was reinventing computing. They were inventing what is now known as cloud computing. That felt very special, so I decided to go in as an engineer. Little did I know that I was going to join a business that was going to be among the largest that the planet has ever produced.

That was the ads team at Google. I went from being an IC [individual contributor] to leading a team of over 10,000 people and making over $100 billion for the company. I took lessons in leadership, humility, scale, and relentless drive from that role.

I learned from many great leaders. What looked like a placid business from the outside had an enormous amount of churn from a growth perspective. For example, making the mobile transition was among the most terrifying things that we did, because it was not clear that we would “make it” back in 2009.

Query and growth had flattened, and at one point, everyone was predicting the end of Google as a company. I had the privilege of learning from the relentless wisdom of the founders, Larry Page and Sergey Brin, but also the thoughtful leadership from people like [former Google CEO] Eric Schmidt.

One of Larry’s maxims was, “Never aspire to be someone successful. You’ll always fail.” His point was that just saying you want to be some other successful company rarely works. By the time you get there, the target has moved. He was always a big fan of finding your own reason to succeed and bringing your own imprint on something. If you don’t have that imprint, perhaps you shouldn’t be doing it.

Looking back, Neeva AI was an impetuous, wild experiment. I wanted to reimagine search. There’s nothing wrong with trying to do that, but I don’t think I had the right insights. So, as it sometimes happens in life, you begin a journey and hope you find the way. Neeva did not.

We built a very competent product, but making it a critical success was really hard. We had an amazing team and amazing tech. My cofounder and I concluded that we would be a lot more helpful to Snowflake, which is why the acquisition happened. Snowflake has been an amazing journey. I’ve worked in data all my life. Among other things, I’m a reformed academic. I did academic research for ten-plus years in databases and query processing. I felt very much at home at Snowflake.

Grit and adaptability outrank any career plan

Eric Kutcher: If you could go back 20 years, with all the wisdom that has come from your career, what’s one piece of advice you would offer yourself?

Sridhar Ramaswamy: Be a little more grateful for the moment, which I try to practice today.

Eric Kutcher: It’s so true. Many young undergraduates join our firm, and they all try to figure out what they want to do next. Do they go back to business school? Do they do something else? A number of these students expect me to say, “Of course you should stay doing what you’re doing.” I think life is about the journey, and there is no such thing as a destination.

Sridhar Ramaswamy: I cannot agree more.

Eric Kutcher: If you were talking to someone who wanted to be in your shoes one day, what advice can you offer as they prepare themselves? I realize the thing you always hear is “no one is ever ready until you’re in those shoes.” But what are the two or three things that you would tell them are important to think through before stepping into a role like yours?

Sridhar Ramaswamy: I don’t think you plan for something like this. I believe in the virtue of hard work and that it’s a requirement for greatness. Some people are magically geniuses, God bless them. But most of us earn it the hard way by putting in the hours and the intensity. I am who I am because I don’t get tired. I put in the energy. I tell people I work seven days a week, and I’m happy to do it. It’s how I trained myself. Hard work is important. That’s number one.

Number two: Malleability is important. The world is ever-changing, especially at a moment like this. You must be able to adapt your beliefs as things change and have the ability to change yourself as you go along. A question I ask people when they want a senior job at Snowflake is, “Tell me something meaningful that you have changed in yourself over the past year.” It can’t be something small, but it has to be meaningful.

The third piece of advice would be not to set limits for yourself. I’m not saying to be greedy or aspire to be a gazillionaire. I mean that you should not have preset limits on what you can and cannot do. Yes, I’m an engineer by training, but I am perfectly capable of everything from communication and marketing to strategy and sales. Obviously, we have experts who have a lot more wisdom. However, being genuinely curious opened up many opportunities for me.