How Intuit plans to ride out the ‘SaaS-pocalypse’

Can you “vibe code” your own QuickBooks or TurboTax?

Go ahead, Alex Balazs said. He’s the CTO of Intuit, the parent company of those financial tools. If a business builds interfaces with any of the major LLM services, they can still tap Intuit’s tools for tasks like invoicing or tax services through integrations with those AI platforms.

Deals with the likes of Anthropic and OpenAI are part of how Intuit is rethinking its business for a post-SaaS-pocalyptic world. Like other business software platforms, Intuit’s stock has cratered this year amid an investor sell-off spurred by advances in coding and agentic AI tools.

We spoke with Balazs about what he thinks this narrative gets wrong, how Intuit is testing new pricing models, and how to market tax services to AI agents.

This conversation has been edited for length and clarity.

What’s your take on the SaaS-pocalypse narrative?

If you look at SaaS products, there’s two layers of them. The first is UI and workflow, and then the second layer, the layer below, is capability and data. And when you get into capability and data, it includes not just data you can get today, but historical data, historical business logic, domain logic, experience building that domain logic. And then on the top is the UI and the workflow. Most people, when they look at the SaaS world, all they see is the UI and the workflow. And UI and workflow is something that large language models absolutely are disrupting in the way we think about it.

The SaaS-pocalypse has kind of cast one brush across the entire industry, when the truth is there are some companies that have made their money, that have made their mark because of UI and workflow, and they’re likely to be disrupted. And there’s a second class of companies that have been UI and workflow, but also they have great capability, great data, great domain knowledge. And those are the ones that if they do it right, they’re going to thrive in this world.

What are some examples of this data and domain expertise that someone vibe coding their own Intuit products wouldn’t have?

Given enough time, I think anyone could vibe code an interface that delivers a simplified experience for a unit of one. And they could probably do so better than any SaaS company out there, because the SaaS companies build UI for millions of customers, not for one.

That’s an important thing when you’re thinking about vibe coding—that vibe coding tends to be built for a unit of one. That’s the reason why, for us, we’re actually happy to show up in marketplaces like Anthropic’s Claude Cowork. Because you can imagine a customer going to Cowork—they’re a brand new small business, and they say, “Hey, I’m a new business. I do lawn care, and I’ve got five customers. I have no idea how to invoice them or collect payment. Can you help me create an app to do that?”

And what Cowork will do is it will build the UI, it’ll build the workflow, but it will call services and tools at Intuit, say, “Hey, here’s the Intuit invoicing service, here’s the Intuit CRM service, here’s the Intuit payment collection service,” and it will wire all these services in. So it’s less about, could someone replicate a UI in a vibe-coding experience or a Cowork experience. They can! And for many customers who are new small businesses, they’re simple tax filers, that’s a very reasonable thing to do. As complexity goes up, that’s when there’s a moment where you’re probably better off coming over to a workflow and a UI that is actually at the SaaS company itself, and that’s what we’re seeing: We’re seeing young businesses going to a place like OpenAI’s marketplace. They start doing basic things, and then once they actually want to do more complicated management of their business, they click a button and they actually come over to us and they sign up for QuickBooks.

Are you rethinking how pricing works for this new agentic era?

Absolutely, yeah. Any company that is just purely thinking about the old-world SaaS subscription model is going to have trouble continuing to grow. We are looking at various types of monetization models, both inside of our own platform and off-platform as well. We’re actually testing things on our own platform, where you use some of these AI features, and the first five times you use it, it’s free. And then by the sixth time, we say, “Hey, you need to buy some tokens or you need to do some additional kind of monetization mechanism.” And we’re learning what works and what doesn’t work. Ultimately, customers will pay when they feel value. And the problem with subscription models is you don’t really always know, am I getting value out of this monthly thing that I’m just paying for? It’s the difference between paying 18 bucks a month for Netflix versus just buying a movie and streaming it online. And if you rent one or two movies, you’re probably better off consumption-based. If you do it a lot, you’re better off subscription-based. And so we think both models will persist in the future, but you definitely need two models, and we’re actively testing those.

So you are rethinking platform interfaces and pricing. How else are you preparing for this new AI era?

Historically, we’ve had product-led growth. You create a great product, you can do marketing, and then people show up and they buy your product. You can have sales-led growth, where a human sales person calls out and says, “Hey, you should buy this product and this is why,” and they help you sign up. [Now,] this is agent-led growth, where the agent basically says, “I need to build something for this customer, and I’m going to make the choice, and I’m going to do the buying for them.” And so we’ve spent a lot of time and energy really understanding—through inspection, through some third-party tools that we work with, and by working directly with the LLM companies—why does an LLM pick one tool versus another tool? And we have all kinds of dashboards that we’re constantly tracking. If you ask this LLM the question, do we show up?…We know why it picks certain tools versus another. [For] certain LLMs, video content is really important. Other ones, content on Reddit is really important. Other ones, content on your own website is more important. So we’re aware, for the three major LLMs, why do they pick one versus another? And we explicitly manage this as a program to make sure that we’re basically marketing to the LLM so the LLM finds us, so it’s the combination of those things.