Three macro trends shaping SaaS and AI in 2026

Executive overview

Public SaaS stocks are falling while AI tools boom — yet the fundamentals of software haven't changed. Companies with proprietary data, recurring revenue models, and the fewest clicks to a result will win.

The three trends are not new: they mirror every prior platform shift (on-prem to cloud, cloud to mobile). The winners play to the shift rather than panic about it.

Distribution and go-to-market remain the hardest, most durable moat — AI cannot replicate them.

Trend 1: Proprietary data is the real asset

• SaaS platforms fall into three tiers by value: system of record (CRM, ERP, HRM — highest), system of engagement (where work happens), system of decision (analytics — lowest).
• AI wrappers that sit at the decision layer added a chatbot and called it AI. They gained little advantage.
• Agents operating in the engagement layer are the real disruption threat — but they fail without access to the system of record's data.
• A BDR agent trained only on LLM data produces generic output; one trained on a company's CRM history produces results.
• Thin wrappers around LLMs will die unless their proprietary data or prompting produces results measurably better than typing directly into the model.
• Existing SaaS companies with data have an under-appreciated moat; they should add agentic features, not defensive chatbots.
• Startups will attempt to integrate into incumbents' data; incumbents will wall off that data or acquire the startups.

Trend 2: Recurring revenue stronger than ever

• Over 75% of OpenAI's revenue comes from subscription (seat-based) pricing — the SaaS model, not usage spikes.
• The business model of SaaS — predictable recurring charges — is valued more highly than one-time revenue and remains the preferred structure for VCs and PE.
• Seat-based pricing is evolving: a platform fee plus credits for agentic workload is the emerging pattern.
• Enterprises want spend predictability; pre-purchasing credit blocks (like AWS reserved instances) lets them model ROI before committing.
• Token costs are falling fast, improving unit economics for AI SaaS.
• AI companies are targeting the labor market (replacing headcount), not just the software market — credits scale with work done, not seats filled.
• Net expansion in AI SaaS comes from credits consumed as agents do more work, not from hiring more users.

Trend 3: Fewer clicks to aha

• Users buy software for a result, not for software itself. Every reduction in clicks to that result is a competitive advantage.
• In traditional SaaS, reaching an aha moment could take 10–100 clicks; getting a full result, 1,000.
• AI compresses this: what took 100 clicks can now take one. What took 1,000 can take two.
• This has always driven competition between SaaS vendors; AI amplifies it dramatically.
• Fewer clicks only delivers value if the result is accurate — which loops back to proprietary data.
• The three trends reinforce each other: proprietary data improves results, better results justify recurring pricing, and agents deliver results in fewer steps.

Why these trends hold: the platform shift pattern

• Three shifts have happened: on-premise to cloud, cloud to mobile, and now cloud to AI.
• 49–85% of global IT spend is still on on-premise software — two full platform shifts later. Disruption is slow.
• "Mobile first" was a differentiator; now it is assumed. "AI-powered" is following the same arc.
• Incumbent response follows a predictable sequence: build in-house (usually fails), then acquire, while independent winners also emerge.
• SaaS stocks falling reflects re-rating anxiety, not immediate revenue collapse.

How to win

• Stop debating whether you are a SaaS or AI company; focus on the customer problem.
• Acquire proprietary data — start by integrating (borrowing), then move toward owning it before incumbents wall it off.
• Price for expansion: platform fee plus credits ties revenue to value delivered.
• Set fewer clicks to aha as the North Star product metric.
• Nail distribution before product perfection: in a world where anyone can vibe-code features, go-to-market is the durable moat.
• AI cannot yet replace human taste and judgment in zero-to-10M go-to-market execution.