gshc2020.com

Pundits dubbed this event the "SaaSpocalypse," with some making bold predictions that it would be the end of software. The logic, borrowed from Sequoia Capital's thesis that AI "autopilots" would replace software tools the way autopilots replace pilots, was simple: If an agent can do the work, why would anyone keep paying for the software that once helped a human do it?

Seven months later, the panic looks premature. The iShares Expanded Tech-Software Sector ETF (ticker: IGV), a widely watched barometer for the enterprise SaaS industry, tells the story better than any single company's earnings call. After opening 2026 near $106, the fund cratered to a 52-week low below $75 in April, a decline of roughly 30%, pricing something close to a permanent SaaS devaluation. By mid-September, it had clawed almost all the way back, trading around $105, essentially flat for the year.

None of this means the AI-agent disruption story was wrong. It means the story got the target backward. Applications aren't disappearing. Their job is simply changing.

From front door to back office

For two decades, enterprise software competed on the interface: the dashboard, the workflow builder, the mobile app that got a worker to log in one more time a day. That was the product. Increasingly, it isn't. Knowledge workers are doing more of their day-to-day work by typing into an LLM chat window and letting an agent fetch what it needs, whether that's a customer record in Salesforce, a head count plan in Workday, or a ticket queue in ServiceNow. Gartner expects 40% of enterprise applications to ship with task-specific AI agents by the end of this year, up from under 5% a year ago. Analysts at A16z have described the shift more bluntly: The traditional system of record "slips into the background as a commodity persistence tier," while a dynamic agent layer becomes the thing people actually see and touch.

The vendors that got hit hardest in February seem to agree, and most were already repositioning as agentic-first before the sell-off forced the issue. Put together, this is not evidence that legacy SaaS is going away. It's evidence that they are becoming less visible. The systems of record that used to be the enterprise's front door are turning into plumbing: still essential, still holding the data of record, but increasingly encountered by employees only indirectly, through whatever chat or agent interface that now sits on top of the stack.

Why the real risk was never the app layer

Here is the more useful debate enterprises should be having instead of "which app dies next": If the interface is moving to chat and the work is moving to agents, what happens when the data underneath isn't ready for either?

The evidence says most companies aren't ready, and that this — not model quality — is the actual bottleneck holding agentic AI back. A pulse survey by Harvard Business Review Analytic Services, sponsored by my company, Reltio, found that 94% of organizations are exploring or implementing AI, but only 15% consider their data foundation "very ready" for agentic AI, and just 39% call themselves highly proficient at ensuring data is trustworthy, even though 94% of leaders rank that trust as their single most critical capability. Gartner predicts more than 40% of agentic AI projects will be canceled by the end of 2027, citing unclear value and inadequate governance over the mounting cost of scaling on shaky foundations. A 2025 MIT NANDA study, still one of the most-cited data points on enterprise AI's return on investment, found that 95% of enterprise generative AI pilots had so far produced no measurable financial return. Informatica's 2026 CDO survey puts a finer point on it: Among companies that have already adopted agentic AI, half say data quality and retrieval are the single biggest barriers to getting anything into production.

An AI agent that acts on — not just answers — bad, stale, or siloed data doesn't just produce a wrong answer. It takes a wrong, and sometimes irreversible, action.

Stop rebuilding the app. Fix what feeds it.

This is the strategic mistake enterprises risk making in reaction to the SaaSpocalypse headlines: pouring budget into recreating Salesforce- or Workday-like functionality in-house, on the theory that owning the software means owning the future. It's the wrong fight. The value was never really the login screen. It was the customer, product, supplier, and employee data that those systems accumulated over many years, and whether an AI agent can trust that data enough to act on it safely. Data is every organization's competitive moat; tools are commodities.

That argues for enterprises spending less energy re-platforming applications knowledge workers barely open anymore, and more on building the data foundation that lets whichever interface wins, chat, agent, or otherwise, actually work: unifying records across the silos that 46% of leaders still cite as their top barrier, closing the governance gap that leaves most companies "highly proficient" in name only, and giving agents a real-time, connected, semantic understanding of the business, not just access to more raw data.

The SaaSpocalypse made for a dramatic six weeks in the market. But the more consequential shift is quieter: The systems of record enterprises built over the last twenty years aren't dying, they're retreating into the background, and the winners in the agentic era will be whoever makes sure what's back there is something an AI can actually trust.

Explore how Reltio connects trusted data for the agentic AI era.

This sponsored post was supplied by Reltio, an SAP company.

• Enterprise Software
• Tech
• AI
• More
• Studios Enterprise