Is AI killing SaaS or exposing weak products?

Why this matters

The loudest voices in tech — including people actively building the replacements — are saying traditional SaaS is dead. Vibe coding has lowered the barrier to building internal tools. One agent can now do the work that previously needed five people clicking around in five products. Per-seat pricing is wobbling. If you lead or advise a SaaS company, you need to know what actually changes and what does not — and have a clear answer to the question any smart buyer will eventually ask: why would I pay for this when AI can do it itself?

Key ideas

The SaaSpocalypse argument is not wrong — it just does not apply equally. Building software is getting fundamentally easier. Seat-based pricing is under pressure from agents. And yes, some SaaS products were never defensible — feature bloat, unclear value and customers renewing out of inertia rather than outcomes. Those are at risk. But "some SaaS companies are vulnerable" and "all SaaS is dead" are very different statements, and collapsing them leads to bad decisions in both directions.

The core argument for SaaS dying falls down on systems of record. Businesses still need permissions, governance, compliance, auditability, repeatable workflows and board-level reporting. They still need cross-functional accountability and a shared view of what is true. Predictable, if-then logic is not glamorous — but when a forecast is going to a board, when a compliance workflow has legal consequences, when a renewal commitment affects revenue visibility, leaders want deterministic controls. AI handles ambiguity better than logic-driven systems. You still need both.

AI is eating shallow admin first. Manual note-taking, activity logging, basic lead scoring, basic health scores, simple routing — that is already happening. Much of what CRM asked people to do was painful, resented and — as Jason Whitehead put it — held together by "indifference and caffeine." AI is actually better suited to those jobs than humans ever were. What AI is not consuming is the layer that holds accountability, governance and a shared operational truth. The interface may go away. The operating backbone underneath does not.

Not all SaaS gets hit the same way. A thin workflow product with weak differentiation and one or two teams using it is in a very different position from a deeply embedded enterprise platform with governance, approval workflows and multiple teams depending on it. The question is how important your product is to how the organisation actually creates value — and whether customers think of it as infrastructure or as a convenience. If it is closer to convenience, that is a risk worth naming now rather than in two years.

The real problem is still adoption — not capability. AI has the same fundamental challenge CRM had for twenty years. The system works. The value is there in theory. But nobody redesigned how the business worked, nobody fixed the incentives, nobody built trust in the data. So adoption is patchy and the system underperforms. If the underlying process is broken, an AI agent will run it at scale — and expose the problem faster, not solve it. Some humans can work around a bad process. Agents cannot. Deploying AI into an immature process and calling it transformation is a specific risk the Jasons name directly: mistaking automation for control.

Agentic AI creates a new kind of trust challenge. Assistive AI asks people to work alongside a tool. Agentic AI asks people to stop doing something and let the machine do it instead. That is a massive psychological and operational shift. Questions nobody has clean answers to yet: who owns the mistake when an agent gets it wrong? What decisions can it make independently? Where does human review start and stop? You cannot announce that you are moving to agentic without first having clear answers to those questions.

The quality bar is going up — and that is probably healthy. If you cannot explain to an AI-forward buyer why your product deserves to exist, why AI alone cannot deliver the same outcome, you are going to struggle. That is not a crisis for strong products. It is overdue pressure on weak ones. The companies that survive the reset own process, trust, accountability and real workflow depth. The ones that cannot make that case should be thinking now about which parts of their product to sunset — so they can focus on what they can actually defend.

The Breakthrough Challenge

If AI stripped away all the manual admin, the interface polish and the seat-based convenience in your product tomorrow — what would still be left that your customers could not easily replace, and would not want to?

Questions for SaaS leaders

• Which parts of your product are genuinely embedded infrastructure — and which are admin convenience that AI can absorb?
• If your biggest accounts deployed AI agents to handle their day-to-day work, would your product become more valuable to them or less?
• Can you make the case for your product to a buyer who says "why would I pay for this when I can just build it with AI myself?"