The landscape of Software-as-a-Service is undergoing its most volatile shift since the transition from on-premise. While many refer to this as SaaSocalypse, I view it as SaaSConolization, Consolidation and Rationalization. As we move through 2026, the traditional "predictable" SaaS model is being dismantled by the very intelligence vendors rushed to integrate.
Challenge 1
The Rationalization Crisis: Renewals Under the Microscope
For years, "shelfware" was a quiet tax on enterprise budgets. In the current economic climate, that era is over. Finance and Procurement teams are no longer rubber-stamping renewals based on historical seat counts.
Audit Culture
Companies are deploying AI-driven procurement tools to identify idle licenses and overlapping features across their entire software stack.
Rule of 40 vs. ROI
While vendors chase Rule of 40, buyers chase Time-to-Value. If a tool doesn't show a direct line to productivity gains within 90 days, it's cut.
90-Day ROI Gate: What Buyers Are Demanding
Challenge 2
The Cannibalization of the "Seat"
We are witnessing the Seat-Count Crisis. For decades, SaaS revenue was tied to human heads. But as agentic AI begins to do the work of humans, the math no longer adds up.
Chart
The Seat-Count Crisis: Agent Displacement of Human Licenses
Traditional Model (Pre-AI)
10 humans → 10 seats → full revenue
Agentic AI Era
1 agent replaces 10 humans → 90% revenue drop
If an AI agent performs the work of 10 analysts, a seat-based vendor is effectively pricing itself out of the market. Square's 40% workforce reduction is an early signal of this structural shift.
Vendors are caught in a "purgatory" between old-school seat-based models and new consumption-based or outcome-based pricing (e.g., charging "per ticket resolved"). This transition often leads to temporary revenue dips that spook investors.
Chart
SaaS Revenue Model Evolution: The Transition Trough
Seat-Based
Pre-2020
Pay per user per month
Hybrid
2020–2024
Seats + usage add-ons
Purgatory
2025–2026
Caught between models
Consumption / Outcome
2027+
Pay per task resolved
Relative revenue index (100 = peak seat-based revenue). The "purgatory" phase (2025–2026) represents temporary dips as vendors transition pricing models.
Challenge 3
The DIY Threat: Death of the "Feature" Solution
The "moat" is evaporating. LLMs have turned complex coding into a commodity. Many SaaS companies are essentially "wrappers" around a specific workflow, features that are now DIY-feasible for internal IT teams.
Shadow Engineering
Using AI coding assistants, non-technical teams are building internal "mini-apps" that replicate 80% of a vendor's functionality at a fraction of the cost.
Feature vs. Platform
If your product is just a "feature", a specific reporting dashboard or basic automation flow, you're now competing with an internal script that takes an afternoon to write.
Chart
Feature vs. Platform Positioning Matrix
- Reporting dashboards
- Basic automation
- Single-workflow tools
Easy to DIY with AI coding assistants. High risk of displacement.
- Salesforce
- Workday
- ServiceNow
Deep data moats, complex integrations, and proprietary workflows survive.
- Niche analytics
- Point integrations
- Single-API wrappers
Low defensibility, low uniqueness, sunset candidates.
- Intercom
- HubSpot
- Zendesk
Racing to embed AI deeply before LLMs eat their interface layer.
Challenge 4
Ceding Control: The Era of "Headless" SaaS
The most existential threat is the loss of the UI moat. LLMs are becoming the "primary interface," turning once-mighty SaaS platforms into mere back-end data repositories.
The Headless SaaS Stack
User Interaction Layer
Claude, GPT, Gemini
Workflow Orchestration
Agentic frameworks
SaaS Application Layer
Intuit, Salesforce
Data Repository
Proprietary records
Infrastructure
AWS, Azure, GCP
The Claude-Intuit Paradigm: Anthropic integrates Intuit's financial "brains" (QuickBooks, TurboTax) via the Claude Agent SDK. Users ask "Claude, file my taxes", and Intuit's brand identity fades into the background.
The Bottom Line
To survive 2026, SaaS vendors must move beyond being "systems of record" and become "systems of intelligence", proving they own the proprietary data and specialised workflows that a general-purpose LLM or a DIY script cannot replicate.