AI Isn’t Killing SaaS. SaaS Is Killing Itself.

Adobe Marketo, a $60,000-a-year B2B marketing platform, suffered a day-and-a-half outage that prevented SaaStr from sending its newsletter to 450,000 subscribers, had a broken unsubscribe link for weeks, and then proposed a 20% price increase, according to SaaStr founder Jason Lemkin. Lemkin argues that legacy SaaS is dying not because of AI but because vendors like Adobe have stopped investing in their products, leaving them with APIs that AI agents cannot use, prompting customers to migrate to platforms like Salesforce that offer better agent operability.

Everyone wants the story to be about AI. AI is replacing seats. AI is collapsing pricing. AI is eating the workflow. There’s truth in all of it, and we write about it here every week. But after the month we just had with one of our own vendors, I’m convinced the bigger story is the one nobody wants to tell, because it’s less exciting and more damning: A lot of B2B software isn’t dying because of AI. It’s dying because the people running it stopped caring. Let me tell you about our last few weeks with Adobe Marketo. Then I’ll tell you why it’s the whole story of legacy B2B right now. The Week Adobe Marketo Broke Everything Here is what we paid $60,000 a year for, in 2026, from a $250B company: - Marketo was down for a day and a half. Not degraded. Down. - Because it was down, we didn’t send our newsletter to 450,000+ subscribers one morning. That send is gone. You don’t get a Tuesday back. - The unsubscribe link was broken for over weeks. That is a CAN-SPAM problem on the single most basic function an email platform exists to provide. We could not get it fixed . We got Adobe’s engineering team on the phone. They blamed Salesforce. Then they blamed us. Then they committed to nothing. And on top of all of it, Adobe wants to raise our price 20% next year. Read that last line again. The product lost a day and a half, dropped a 450K send, shipped a broken unsubscribe link for two weeks, and the renewal conversation is a 20% increase. That is not a company fighting to keep a customer. That is a company harvesting one. With borderline contempt for them. The Part That Actually Matters: Old SaaS Can’t Even Be Operated by an Agent We run SaaStr AI with 3 humans and 20+ AI agents. Everything reads and writes through an API. So when Marketo broke, the question wasn’t “can a human fix this.” It was “can our agent fix this.” And the answer was no, because Marketo is not built to be operated by anything except a human clicking slowly through a dashboard. Here’s what our agents ran into trying to do basic work against Marketo this year: No real webhooks. Everything is poll-based. There is no event for an agent to trigger on. Single-digit bulk exports per day. Hit the ceiling and you’re locked out until tomorrow. 90-day log retention. Anything older is simply gone. Auth that breaks on whitespace, silently, with no useful error. No refresh tokens. A Counts API that can’t answer “how many people are on this list right now.“ Six months of history, maximum. A full sync of our own data takes days because of the rate caps. No MCP server. No SDK. No agent toolkit. No llms.txt. Nothing an agent can discover or call. So when the unsubscribe link broke, our agent couldn’t diagnose it, couldn’t file against it, couldn’t push a fix. There was no surface to operate on. We gave up and rebuilt the unsubscribe flow ourselves on Replit in an afternoon. And once you do that once, you start asking the obvious question. We’ve now migrated almost all of Marketo’s functionality onto Salesforce. Not because Salesforce has better features. Because Salesforce has an API our agents can actually use. One day you look up and realize you’re paying $60K a year for a product you’ve already routed around. That is the new B2B churn motion. It isn’t a feature bake-off. It’s an agent-operability test, and legacy vendors are failing it quietly, one migrated workflow at a time, before they ever show up on the churn report. Adobe is still pricing and selling Marketo like it’s 2019: built for a world where big ops teams click through screens and escalate to humans at the vendor. That world is ending. It already ended at SaaStr. This Is the Whole Story of B2B Right Now Here’s the thing. AI didn’t break Marketo’s API. Adobe did, by not touching it for years . AI didn’t drop our newsletter send. Downtime did. AI didn’t propose a 20% increase on a degrading product. A spreadsheet did. AI is the thing that finally made all of it visible. When your agents can’t operate a product, the product’s neglect stops being a minor annoyance and becomes the reason you leave. And once you see the pattern, you see it everywhere. There are a few distinct ways B2B companies are killing themselves right now, and not one of them is “AI got too good.” 1. They stopped modernizing, and the AI era exposed it. Marketo is the clean example, but it’s a category. A product architected in 2015 with a human-first API, layered with a few AI features bolted on top, is not an AI product. The market can tell the difference between AI-native and AI-added, and so can your customers’ agents. The first time an agent fails to operate your software, you’ve started losing the account. You just won’t see it on a report for two more quarters. 2. They raised prices on a worse product. The most telling line in the whole Marketo story isn’t the downtime. It’s the 20%. When a vendor’s answer to declining quality is a price increase, that’s not confidence. It’s extraction. It’s a company that has quietly decided to milk its base instead of earn it, which works right up until the moment switching cost stops being a moat, and AI is busy demolishing switching costs. 3. They’re managing to a debt number, not a product. Look at the PE-backed software wreckage of the last 18 months. Pluralsight’s equity was wiped in 2024. Medallia’s $5B+ in equity went to zero in 2026 when Blackstone refused to extend the PIK window. Proofpoint, Qualtrics, Quest, and Cornerstone are on every distressed-credit watchlist. These companies aren’t dying because a startup out-shipped them. They’re dying because they were loaded with debt at peak-2021 multiples, and now every dollar goes to debt service instead of the roadmap. When you’re managing to a coverage ratio, you don’t rebuild your API. You don’t fix the unsubscribe link. You raise the price 20%. 4. They stalled, and decided it was the market’s fault. A year ago we ran the numbers on the public B2B market, and they haven’t gotten kinder. Companies growing 20%+ trade at roughly 12x. Companies growing under 10% trade at roughly 3x. That’s a ~80% valuation discount for slipping below the growth line. There’s a whole cohort, Yext, Domo, LivePerson, Upland, trading below 2.5x revenue, several below 1x. The 2x ARR Club . Every one of those companies will tell you the market is being irrational. The market is not being irrational. The market is telling them it doesn’t believe their future is bigger than their present. And mostly, it’s right. None of those four is AI killing SaaS. All four are SaaS killing itself, with AI holding the mirror. What This Means for B2B If you build or run a B2B company, the Marketo story is a checklist of what not to become. Make your product operable by an agent, now. Not next year. The question is no longer “is our UI good.” It’s “can a customer’s agent do real work against our API without a human babysitting it.” If the answer is no, you are already on a churn clock you can’t see. Webhooks, an SDK, an MCP server, clean auth, real history. This is table stakes in 2026. Never let your renewal motion get ahead of your product. A price increase on a degrading product is the single clearest signal that a vendor has given up on earning the relationship. Customers feel it. Your best ones leave first, because they’re the ones already capable of building the replacement. Watch your capital structure like it’s a product feature. If you’re carrying peak-vintage debt, you will be tempted to manage to the coverage ratio and let the product rot. That’s the trap that wiped Pluralsight and Medallia. Cash buys time, and time is what you need to make the AI transition. Debt takes it away. Grow, or be honest that you’ve stopped. The 60% discount for sub-10% growth isn’t a public-market quirk. It’s a preview of what happens to any B2B company, public or private, that decides “steady” is a strategy. Re-accelerate or get genuinely, deeply profitable. The middle is the most dangerous place to be. We didn’t leave Marketo because a competitor dazzled us. We left because Adobe stopped showing up, and our agents had somewhere better to go. That’s how this ends for a lot of B2B software. Not with a dramatic AI death. With a slow, self-inflicted one, and a 20% price increase on the way out. AI isn’t killing B2B SaaS. It’s just making it impossible to hide that some of it was already dead. Data approximate and point-in-time as of mid-2026. Not investment advice.