Last week, $830 billion in market cap vanished from software stocks. The trigger? Anthropic released legal-focused plugins for Claude Cowork, and suddenly everyone was convinced that AI agents could horizontally replace knowledge work. Thomson Reuters dropped 16% in a single day. The term "SaaSpocalypse" started trending.
At this week's AI Impact Summit in New Delhi, Salesforce executives pushed back hard against the doom narrative. "SaaS is dead" is counterfactual, they argued—even leading AI model developers rely on enterprise platforms for mission-critical functions.
So who's right? Are we witnessing the death of software, or just another cycle of tech industry hysteria?
The answer is more nuanced—and more interesting—than either camp admits. And if you're a product manager, the implications for how you think about customer insights are profound.
The Three-Layer Reality Check
Here's the framework that actually makes sense of this debate: SaaS isn't one thing. It's three layers stacked on top of each other, and AI agents are disrupting each layer very differently.
The UI Layer is what users click on. Dashboards, input forms, report views. This is the part that's genuinely threatened. When an AI agent can extract deal information from emails and calendar events and register it directly in your CRM, you don't need a sales rep clicking through screens to do data entry. The human-operated GUI becomes unnecessary.
The Workflow Layer handles approvals, business rules, and task routing. This is where competition intensifies. Yes, AI agents make building workflows faster. But building and operating are different problems. Who keeps the audit trails? Who does root cause analysis during incidents? Who manages access controls? The real question isn't construction cost—it's who takes operational responsibility.
The System of Record (SoR) Layer is where the actual data lives. Accounting, contracts, customer master data, audit logs. This layer doesn't just survive—it becomes more important. Fortune 500 companies can't easily switch systems that have accumulated decades of data integrity, integration testing, and audit compliance. What enterprises pay SaaS vendors for isn't just code—it's the price of risk transfer.
Analysts at SmartScope put it bluntly: "The February 2026 SaaS stock crash was the market overreacting to an 'AI agents fully replace SaaS' narrative." When someone says "SaaS is dead," if they're talking about the UI layer, they're largely right. If they're talking about the SoR layer, they're off target.
Where Customer Insights Fit in This Picture
Here's what nobody in the "SaaS is dead" conversation seems to be asking: What about the voice of your customer?
Customer insights—feedback from support tickets, interview transcripts, survey responses, product reviews, churn conversations—sit firmly in the System of Record layer. This is foundational data that informs every product decision. It's not a dashboard to be replaced or a workflow to be automated away.
If anything, the compression of UI and workflow layers makes customer insights more critical, not less. Here's why:
When AI agents build features faster, knowing what to build becomes the bottleneck. If your engineering team can ship in days what used to take weeks, the constraint isn't execution anymore. It's judgment. Product intuition. Understanding what customers actually need versus what they say they want. Getting this wrong at AI speed means you'll fail faster and more expensively.
When everyone has access to the same AI development tools, competitive advantage shifts to insight quality. If your competitors can also spin up features with AI agents, the differentiator isn't development velocity. It's whether you're building the right things. The teams with the clearest understanding of customer problems—distilled from hundreds of conversations, not hunches—will win.
When per-seat pricing dies, customer retention becomes everything. Salesforce has already started transitioning to flat-rate AI licensing (they call it ALEA—Agentic License Enterprise Agreement). When your revenue model isn't tied to seat count, you need customers who stay, expand, and advocate. That requires actually understanding what makes them successful. Which requires—you guessed it—customer insights.
The Practical Playbook for Product Teams
If you're a PM navigating this landscape, here's what actually matters:
1. Audit Your Insight Infrastructure
Ask yourself: Where does customer feedback actually live in your organization? If the answer involves phrases like "scattered across Notion docs," "trapped in sales call recordings nobody watches," or "somewhere in that Slack channel," you have a SoR problem masquerading as a process problem.
The teams that thrive in the agent era will have their customer insights consolidated, searchable, and connected. Not because it's nice to have—because when AI agents can execute on product decisions in hours instead of weeks, you can't afford to spend days hunting for the customer context you need.
2. Separate Signal from Noise Before AI Amplifies Both
AI agents are incredibly good at pattern recognition. They're also incredibly good at pattern recognition on garbage data. Feed an agent your unfiltered support ticket history and you'll get confident-sounding insights that reflect your support workflows more than your customer needs.
The preprocessing work—categorizing feedback, identifying themes, separating feature requests from complaints from confusion—isn't grunt work that AI eliminates. It's the foundation that determines whether AI helps you or leads you astray.
3. Build Your "Why" Repository
Every feature shipped solves a problem. Every problem came from somewhere—a customer interview, a churned account, a support escalation, a competitive loss.
Document the chain. When AI agents ask "what should we build next?" (and they will, increasingly), the answer should trace back to real customer evidence. "We're building X because customers Y and Z experienced problem W, and our research shows this affects N% of our ICP." Not "we're building X because it seemed like a good idea."
The companies that connect decisions to customer evidence will make better choices. The companies that don't will make faster bad choices.
4. Treat Churn Conversations as Gold
In a world where customer acquisition gets more expensive and net revenue retention becomes the primary growth metric, every churned customer represents both a learning opportunity and a failure to learn earlier.
The friction points that led to that churn conversation? They existed for months in support tickets, usage patterns, and NPS comments. The product teams that surface these signals before the churn call—not after—will retain customers their competitors lose.
The Real Lesson from the SaaSpocalypse
Here's what the market panic of February 2026 actually reveals: we're still terrible at distinguishing between interface disruption and value disruption.
AI agents will absolutely replace dashboards and data entry screens. They'll compress workflows and automate approval chains. That's interface disruption—and it's real and significant.
But the underlying questions that product teams need to answer haven't changed: What do our customers actually need? Why are they churning? What features would make them successful? Which problems are we solving, and for whom?
These questions don't have AI-generated answers. They have customer-generated answers that AI can help you find faster—if you've built the infrastructure to capture and organize that customer voice in the first place.
The companies that confuse "AI can build features faster" with "AI knows what features to build" will discover the difference painfully. The companies that invest in understanding their customers deeply—and connecting that understanding to their product decisions systematically—will use AI to accelerate good judgment instead of automating bad judgment.
SaaS as an interface paradigm may be dying. But SaaS as a delivery model for software that solves real customer problems? That's evolving, not disappearing.
And the teams that know their customers best will be the ones who evolve with it.
The voice of your customer shouldn't be trapped in scattered tools and forgotten documents. Pelin helps product teams turn customer conversations into actionable insights—automatically surfacing the patterns that matter before they become churn conversations.