By Kevin Keenan, vice president of communications at Reltio
Software stocks did not fall 23% in six weeks because investors suddenly lost interest in recurring revenue. They fell because Anthropic made it easy for anyone to build a competent digital worker.
For two decades, applications have been the front door to work. You log into a CRM, ERP, or HCM. You search for a record. You interpret the data. You move it somewhere else. Each app owns its workflow, its interface, and, effectively, its slice of truth.
Anthropic's breakthrough did something subtle but profound. It suggested that the default unit of work may no longer be the app. It may be the agent. And if agents can orchestrate work across systems, the apps' monopoly on how work begins starts to erode.
The interface is flipping
Enterprise software has been dashboard-driven. It was designed for humans navigating menus, forms, and permissions. But agents don't care about dashboards; they care about outcomes.
Instead of "Open Salesforce and update the opportunity," the instruction becomes: "Handle this renewal risk." The agent gathers context, checks permissions, executes the necessary actions across systems, and logs its actions.
Software does not disappear. Finance still closes the books in the ERP. HR still manages onboarding in the HCM. Systems of record remain critical. What changes is the interface. The conversational layer becomes the entry point, and applications recede into infrastructure.
Markets are forward-looking. Over the past year, software has underperformed the S&P 500 because investors expect this shift. If agents can replicate and coordinate what users once did manually inside apps, then the value pool moves. The question is: to where?
The context gap
The excitement around agents hits a hard wall in most enterprises: messy data.
A recent Harvard Business Review Analytic Services pulse survey, Unlocking the Data Advantage in the Age of Intelligence, makes the readiness gap clear. While 93% of organizations are exploring or implementing AI, only 15% describe their data foundation as "very ready" for it. Nearly half of leaders cite data silos as their biggest obstacle to making enterprise data usable for AI.
This is not a small technical issue; it's existential. When truth is fragmented across disconnected systems, an agent becomes a source of frequent errors, rather than the productivity multiplier it's meant to be. It can act quickly, but not wisely.
Anthropic's models made agents capable. Enterprise data makes them dangerous or powerful. That tension is at the heart of the software crash.
The new moat is context, not UI
If agents become the new front door to work, a durable advantage no longer sits in the prettiest dashboard or the stickiest workflow. It sits in the layer that supplies trusted context and governs execution.
In the HBR survey, Manish Sood, CEO and founder of Reltio, states that "unified, real-time, trustworthy data is the context that powers the shift to agentic AI." Without a unified context, agents lack situational awareness. They cannot reliably answer basic questions: Is this data current? Is it authoritative? Am I allowed to act on it? Will my action conflict with another system?
This has elevated a new requirement in the enterprise stack: a context layer. Its job is simple to describe and difficult to execute. It must unify meaning across silos and enforce two non-negotiables:
- Right now: The data must be current, deduplicated, and reconciled.
- Allowed: Permissions, governance, and auditability must be embedded in every action.
Most high-value workflows — such as resolving a supply chain disruption, flagging a renewal at risk, or identifying fraud — require coordination across CRM, support, billing, procurement, and inventory systems. They do not live inside a single application.
Investors understand this. Many incumbent SaaS platforms are optimized to own a domain, not to harmonize truth across domains. Agents expose that limitation immediately.
The execution gap
The HBR research highlights why markets are punishing "wait-and-see" vendors.
While 94% of leaders say trust in data reliability is essential for AI success, only 39% consider their organizations highly proficient in it. Eighty-nine percent value strong data governance; just 37% believe they have mastered it. It's clear that the gap between aspiration and execution is wide.
"Data governance is 100% a strategic differentiator," notes Robbie Beyer, director of data science at RSM, in the HBR survey. "If you invest in a resilient, evergreen data platform, it becomes more valuable over time."
In an agentic world, this is no longer a back-office discipline. It becomes the foundation of competitive advantage. Markets are repricing software because the value of a stand-alone workflow shrinks when orchestration moves up a layer. The companies that survive the SaaS correction will not be those that bolt an agent onto a legacy stack. They will be those who reorganize around context.
Who owns truth, permission, and action?
The real question behind the 2026 sell-off is not whether SaaS dies. It is who owns the hardest parts of the workflow in an agent-driven enterprise. Is it the model? The application? Or the layer that guarantees truth, enforces permission, and logs every action?
We do not lack powerful LLMs. Anthropic, OpenAI, and others have made them abundant. What remains scarce is a trusted, permissioned context with provenance: data that is unified, governed, explainable, and defensible.
Ross Schalmo, chief data officer at Eaton, summarizes it well in the HBR survey that winning organizations are those that "get the plumbing right."
The SaaS-pocalypse is not the end of software. It is the end of software's monopoly on the interface. As the center of gravity shifts from the best UI to the most trustworthy context, valuation will follow.
Agents may be the visible disruptor. But the real reordering of value is happening beneath the surface, in the data foundations that determine whether those agents can be trusted to act. For companies that have unified their data and embedded governance into execution, the agentic era is an accelerator.
For everyone else, the repricing may only be starting.
This post was created by Reltio with Insider Studios.
The post How the 2026 software crash is shifting power from SaaS apps to AI agents appeared first on Business Insider
