Is it really the end of SaaS as we know it?
New AI agent and coding tools have triggered a “SaaS-pocalypse” narrative. The reality is more complicated.
• 7 min read
Are software companies like Oracle and Salesforce headed for the trash icon of history?
Investors have been dumping their stocks in software-as-a-service (SaaS) companies since late January, with one trader (and dozens of headlines) dubbing it the “SaaS-pocalypse.” So while time will tell whether the end is actually nigh, change is in the air. And like seemingly all things business-related in 2026, that change is predicated on some form of AI.
The prevailing theory behind the sell-offs seems to be that sophisticated workplace agents as well as new coding tools from Anthropic and OpenAI could make this sort of business software model obsolete. If an IT team can easily “vibe code” internal software or assign tasks to role-specific plug-ins within Claude Cowork, do they still need expensive pay-per-seat platform subscriptions?
“The thing with agentic systems is now you can have agents just read your business policies and generate the software,” Booz Allen chief technology officer Bill Vass told us. “This is going to disintermediate these SaaS vendors and these ERP vendors. The market just realized that, and that’s the reason you see the big reaction.”
SaaS leaders claim a lot happens behind the scenes of these platforms that their customers can’t easily replicate, such as data governance and security and teams of developers that ensure that software works reliably. But business leaders at companies like Walmart and LiveRamp have told us they’re now building more internal tools and doing so more quickly than ever, if not yet severing SaaS contracts.
What nobody denies is that the SaaS industry is in a period of flux, and companies need to continually evolve to keep offerings competitive.
This sentiment is not exactly new. For instance, when we spoke a year ago with Scott Snyder, senior fellow at the Wharton School, he warned that SaaS companies could struggle to keep up with AI development tools.
“People could say, ‘Well, why am I using a service—[like] a Salesforce customer service bot—when I can just build my own at a very low cost because I’ve already got the AI infrastructure set up?’” Snyder said last March. “If the switching costs aren’t that high, these larger companies may roll out their own, and I think that’s going to be the interesting tension going forward.”
A major evolution
In a way, what comes next could harken back to an earlier model of business software development, according to Vass. Before standardized enterprise resource planning (ERP) software platforms became widespread in the 1990s—a precursor to SaaS—businesses mostly used custom-built internal software, Vass said.
“Originally, we all built custom ERP systems for everyone. So you did it exactly the way the Army wanted…or exactly the way Bank of America wanted, or exactly the way BMW wanted,” Vass said. “Then there was this move that, ‘Well, gee, that’s really expensive to build all those custom systems. Let’s all standardize on SAP or PeopleSoft or Oracle.’…It was very painful to go through that, and every single business person hated it.”
Vass said that the rise of AI agents and coding tools could swing the pendulum back toward more individual customization.
Still, Vass said he thinks many SaaS vendors will likely successfully pivot to selling AI agents, and internal agentic systems haven’t matured to the point where he’s seen companies actually cutting SaaS contracts.
“You’re going to have a combination of people that are going to just build their own agentic, if you like, ERP systems that will still be running on a cloud, probably, but run [on premises], too,” Vass said. “You’ll have others that will say, ‘I’ll just go use Agentforce from Salesforce,’ for example, or something like that, which is an agentic system to do what I do. And then it’s a lot more flexible.”
Those contracts may soon be changing form, however. Faisal Masud, who heads HP’s digital and lifecycle services division within the company’s SaaS business, said that the traditional seat- and usage-based models for pricing SaaS platforms are “evolving.” Masud said HP is mulling an additional option for outcome-based pricing based on metrics like number of tickets eliminated or hardware procurement costs.
Tech news that makes sense of your fast-moving world.
Tech Brew breaks down the biggest tech news, emerging innovations, workplace tools, and cultural trends so you can understand what's new and why it matters.
“We’re considering, as the market evolves, to have both seat-based and outcome-based pricing models,” Masud said.
Enterprise SaaS price increases have been significantly outpacing CIO budgets in recent months, according to Gartner research, with companies often facing 10% to 20% hikes during contract renewals. Gartner expects enterprise software spending to increase by 15% this year, but much of it will go toward price increases and AI infrastructure spend, per the B2B-focused hub SaaStr.
The average annual contract value for private SaaS companies has trended upward in recent years, jumping from a median of $22,357 to $26,265 between 2023 and 2024, according to a survey of more than 1,000 respondents conducted by lending firm SaaS Capital.
Masud said the SaaS market is in for a correction rather than a full-on doomsday scenario. He said smaller point solutions for specialized use cases—“you can guess what those are”—are more likely to be impacted than larger, more integrated platforms. He predicted the biggest effect will be drops in pricing as CIOs negotiate better deals.
“Having built a SaaS company, we were users of SaaS, too…And some of those prices were just completely out of bounds for us and for the economic buyers. Everything was very expensive,” he said.
Back-SaaS
In early February, two CNBC reporters with no coding experience vibe-coded a clone of a dashboard from project-management platform Monday.com in under an hour to demonstrate the threat that SaaS companies face. The result was a working prototype that the reporters said could successfully manage their personal lives. If reporters could vibe-code their way through a dashboard, what then of the multibillion-dollar SaaS companies?
Alexey Korotich, chief product officer at workplace platform Wrike, a competitor of Monday.com, said it’s been easy to build dashboards like that for years—the user interface part of the SaaS offering is already heavily commoditized. What’s harder to replicate, he said, are data governance, back-end infrastructure, and reliability.
“Yes, you can build a whiteboard dashboard in an hour or so, but then, how does that look if you start adding features? How does that look if you start finding bugs here and there?” Korotich said. “You will quickly realize that it’s beyond your capacity to handle. It’s not one person’s job, it’s a team’s job to actually take care of this whole thing.”
Tamar Yehoshua, chief product and AI officer at Atlassian, the company behind platforms like Jira, Rovo, and Trello, said the AI transformation will likely be similar to how cloud services changed software tools, but at an accelerated pace. Cloud services first gave rise to SaaS in the aughts as subscription models replaced traditional perpetual licenses.
“SaaS is not going away,” she said. “People are going to have to build software. People are going to have to communicate with other people. Unless you think that every organization is going to be one person who’s just shooting off an AI.”
Not everyone agrees that the “SaaS-pocalypse” is an overreaction, however. Tatyana Mamut, a former Salesforce VP who now runs a startup called Wayfound that partners with Salesforce on agent monitoring, expects the SaaS market to majorly shrink over the next half-decade.
This outcome would contradict various market research reports that show the SaaS industry continuing to grow significantly in the next decade amid an ongoing AI transformation.
AI will eliminate the primary switching costs of retraining humans (since much of the work will be done by agents) and restructuring data (because LLMs don't require structured data), she said. Reductions in seats required for most SaaS subscriptions will also eat into profits, she said.
“The way we’re going to know who’s going to be alive in 2030 is to see who launches really great AI agents that are doing full and complete jobs this year in 2026,” Mamut said. “That’s my best bet.”
Tech news that makes sense of your fast-moving world.
Tech Brew breaks down the biggest tech news, emerging innovations, workplace tools, and cultural trends so you can understand what's new and why it matters.