Chris Dalton

The Nasdaq CEO Says AI Won't Kill Software and She's Only Half Right

March 22, 2026

On March 11, 2026, Adena Friedman - CEO of Nasdaq, one of the most powerful financial institutions in the world - sat down with David Rubenstein at an Economic Club of Washington dinner and said out loud what a lot of people in her position have been thinking: AI is not a death knell for software. It's a catalyst.

I want to take her seriously, because she deserves that. She's been running Nasdaq since 2017, she's the first woman to lead a global exchange, and under her watch Nasdaq has quietly transformed from a stock exchange into something closer to a financial technology empire. She's not a commentator. She has real skin in this.

But she's only half right. And the half she's wrong about is the half that matters most to the people actually running businesses right now.

What She Actually Said

Friedman's argument is essentially this: software isn't dying, it's adapting. "Any business that sits still in the world of AI will ultimately fail," she said. That's the through line. Move or die. She pointed to Nasdaq's own AI integration as evidence - including a tool called Settlement Guard, which uses AI to predict settlement failures and help firms avoid losses by catching problems before they happen. Her framing is that AI empowers battle-tested systems when integrated properly, and that the financial industry in particular needs precision and security that only mature, enterprise-grade software can provide.

None of that is wrong. I actually think it's correct. It's also the easier half to be right about.

What Was Happening Around Her When She Said It

Here's the context that makes Friedman's comments more complicated than they sound at face value.

She made these remarks about six weeks after one of the most violent software selloffs in recent memory. On February 3, 2026, Anthropic launched a suite of industry-specific plugins for its Claude Cowork tool, and the markets reacted like the building was on fire. A Goldman Sachs basket of US software stocks dropped 6% in a single day - its biggest one-day decline since the tariff-fueled selloff the previous April. A $285 billion rout hit stocks across software, financial services, and asset management. The iShares Expanded Tech-Software ETF entered a technical bear market. Traders at Jefferies started calling it the "SaaSpocalypse" - a term for the apocalypse-level selloff hitting SaaS companies - and the name stuck. We've written about this phenomenon before on this site, because it genuinely matters.

The panic wasn't irrational. What spooked investors wasn't that Claude could write an email or summarize a document. It was that Claude Cowork's legal plugin was being described as capable of automating the majority of standard NDA and compliance triage. Legal review. Compliance checks. Risk flagging. These are paid workflows - the exact kind of workflows that software companies have been charging per-seat licenses to support for two decades. Within 30 days from mid-January to mid-February, roughly $2 trillion in software market capitalization evaporated.

So when Friedman says AI won't kill software, she's speaking into one of the most severe market credibility crises the software industry has ever faced. That context matters.

The Half She's Right About

She's right that the category doesn't die. I believe that.

The incumbent software CEOs who came out swinging after the SaaSpocalypse made some legitimate points. Workday's founder Aneel Bhusri, brought back as CEO to navigate exactly this kind of moment, said that Workday systems embed two decades of business processes that AI cannot replicate. "AI, for all of its incredible capabilities, is probabilistic by nature," he told analysts. It reasons, predicts, recommends. It is not, yet, a deterministic system. Salesforce, for its part, is sitting on more than 50 trillion real-time records. Oracle's Mike Sicilia said flatly that he doesn't agree with the death-of-SaaS thesis, and that AI tools would only be a threat if companies weren't adopting them - and they are. Even Jensen Huang called the idea that AI replaces software "the most illogical thing in the world."

And the actual revenue data backs this up. ServiceNow's Q4 2025 subscription revenue grew 21% year over year. Salesforce's remaining performance obligations came in at $72.4 billion, up 14%. Snowflake grew 30%. These are not companies whose customers are fleeing. The panic is real; the revenue collapse, so far, is not.

Friedman's argument - that the companies with deep proprietary data and complex, secure workflows are going to be fine - is defensible. It might even be right for the next several years. When she talks about "industrial-strength, secure" software built for financial operations, she's describing something that a general-purpose AI agent actually can't replace tomorrow morning. I'll give her that.

The Half She's Wrong About - Or At Least Not Saying Out Loud

Here's where I think the Nasdaq CEO's framing quietly lets a lot of software businesses off the hook in a way that's going to hurt them.

The threat isn't that enterprise software disappears overnight. The threat is that the per-seat model - the entire economic foundation that the modern SaaS industry was built on - is cracking in ways that may be structural and permanent. When Anthropic's Claude Cowork connects to Slack, Figma, Asana, Monday.com, and a dozen other tools through an open connector architecture, it doesn't kill those products. It just starts to own the workflow layer above them. If you can run multi-step professional tasks across all your software from a single AI interface, you need fewer seats. Maybe not zero seats. But fewer. And fewer seats at scale is a 90% revenue reduction on the same output.

Atlassian already reported its first-ever decline in enterprise seat counts in early 2026. Workday announced an 8.5% workforce reduction, citing AI-driven efficiency gains that paradoxically reduced the need for the very human roles its software was designed to serve. Software price-to-sales ratios have compressed from 9x to 6x - levels not seen since the mid-2010s.

Friedman's version of events - AI empowers software, software gets smarter, everyone benefits - is optimistic in a way that skips over the part where the pricing model for software has to completely reinvent itself. The interface is no longer the value proposition. The outcome is. That's a genuinely hard transition for companies whose entire business was built around selling access to an interface.

She's running Nasdaq, which has proprietary financial data going back decades, and which has products like Settlement Guard that are doing things AI can genuinely enhance rather than replace. She's in the most defensible part of the software market. That's a little like a company that builds nuclear plant software saying AI won't kill software. Well, no - not that software. But she's not running a mid-market CRM or a project management tool or a basic workflow automation product. Those businesses are living in a completely different reality right now, and her reassurances don't quite reach them.

Derek tried to explain this to me using a Star Wars metaphor last week. Something about how Friedman is like the Jedi Council, confident in the institutions, while the real disruption is coming from the edges. I nodded. I don't fully know what that means but it felt accurate.

The Seat Compression Problem Is Real and Nobody Wants to Say It Plainly

Here's what actually happened with the SaaSpocalypse that I think gets underreported: it wasn't really about whether AI can replace software. It was about whether AI can reduce the number of humans who need software licenses. Those are very different things.

If 10 AI agents can do the work of 100 sales reps, you don't need 100 Salesforce seats. You need 10. That's not Salesforce dying - that's Salesforce's revenue model taking a 90% haircut on that account. The software still exists. The workflow still runs through it. The bill just looks completely different. If you're a CRM or a workflow tool or a project management platform priced per seat, this is not a subtle shift in market conditions. This is an existential pricing problem disguised as a technology trend.

AI-powered apps, interestingly, are already showing a version of this pattern at the consumer level. One recent study across more than 115,000 apps covering over $16 billion in revenue found that AI-powered apps generate 41% more revenue per payer - but churn 30% faster. AI sells, but it's not sticking. The initial value is real; the sustained lock-in isn't there yet. That pattern is going to repeat itself at the enterprise level, and the software companies who think "we have a moat" because customers signed contracts last year should be very attentive to what happens at renewal time.

Stephanie sat in on a renewal meeting for a workflow tool last quarter and genuinely couldn't understand why the vendor was nervous. She just assumed everyone would renew forever because they always had. Gerald apparently had a similar assumption about a software vendor Linda uses for something at home. I don't know the full story. The point is, contracts don't renew themselves.

What Friedman Is Right About For Business Owners

I don't want to be unfair to her. There's something practically important in what she said that gets lost in the macro debate.

"Any business that sits still in the world of AI will ultimately fail."

That sentence is correct. And it's more actionable than the death-of-software panic, which mostly just makes people feel bad without telling them what to do. The software businesses that are going to make it through this transition are the ones that have deep proprietary data, complex integrations, and genuinely irreplaceable workflows - exactly the things Friedman is pointing at. The ones that are going to struggle are the ones that are basically a nice interface on top of something AI can now do natively.

There's also a real argument - and the Databricks CEO Ali Ghodsi has made this point well - that the actual threat to incumbent SaaS isn't that AI replaces them outright. It's that AI dramatically lowers the barriers for new competitors. Well-funded startups can now build genuinely competitive products faster and cheaper than ever. The window is open not for individuals to prompt their way to a Salesforce replacement, but for serious companies to challenge markets that seemed impenetrable two years ago. That's a different kind of pressure than "the software dies," but it's not nothing. It might actually be scarier for incumbents than the apocalypse narrative, because it means the competitive threat is coming from everywhere, not just from one AI model.

I spent some time last month trying to set up a lead management workflow using a couple of different tools. I think I had the automation running backwards - it kept triggering on old contacts instead of new ones. I eventually figured out I'd set the filter condition wrong. It took me about 45 minutes. The point being: AI tools are genuinely useful in practice, but the implementation still requires humans who understand what they're actually trying to do. The tools aren't replacing the thinking yet. The seat compression is real, but the complete disappearance of human-guided software work is still a ways off.

The market is already punishing SaaS stocks that can't explain their AI story clearly, and that's not going away. The businesses that will survive this are the ones acting like Friedman's advice is correct - leaning in hard, building actual AI capability, not just adding a chatbot to the UI and calling it done - while also being honest with themselves about whether their pricing model can survive in a world where AI reduces the headcount that buys licenses.

My Actual Take

Adena Friedman is right that software doesn't die. She's wrong - or at least incomplete - in implying that the transformation is orderly and that companies just need to adapt. The adaptation required is more radical than she describes. The per-seat model is in genuine trouble. The interface-as-product model is in genuine trouble. The companies with proprietary data, deep workflows, and real enterprise complexity are going to be fine. The ones that built a business on being the nice way to do something AI now does natively are facing something much harder than a product refresh.

The SaaSpocalypse is a real naming of a real thing - not a permanent death sentence, but a genuine structural reckoning about what software is actually worth when AI starts owning the workflow layer above it. Friedman is describing the world from the vantage point of one of the most defensible software businesses on earth. Most software companies don't have that view.

What I keep coming back to is this: the companies that will be fine are not the ones that believe they'll be fine. They're the ones that are actively, obsessively figuring out what they actually own that AI can't replicate. Proprietary data. Deep integrations. Decades of domain-specific logic baked into a platform. Trust relationships with enterprise customers who can't afford to be wrong. Those are real moats. "We have a good UI and customers like us" is not a moat anymore. It might not have been for a while now.

Friedman said that Nasdaq has 10,000 employees, and about half are in product and technology. That ratio - committing half your workforce to building the thing, not just selling it - is probably closer to the right answer than any individual statement about whether AI kills software. The companies leaning in at that level are going to be okay. The ones hoping the storm passes are not.

Tory told me this week that he's "really excited about the pivot." I don't know exactly which pivot he meant. His car situation hasn't been resolved. He seemed genuinely optimistic, though. I think that counts for something, even if I'm not sure it's enough by itself.

It isn't. But Friedman's half of the argument - move, adapt, integrate, don't sit still - that part is right. I just wish more people saying it were also being honest about what "moving" actually costs.