Where Does SaaS Go From Here? The Next Five Years, Honestly

Software-as-a-service has been the default business model for fifteen years. Build a web app, charge per seat per month, grow ARR, raise on the multiple. That playbook is now under real pressure — not from a competitor, but from a shift in how software gets built and priced.

Here's where we think SaaS is actually heading over the next five years, without the conference-keynote optimism.

1. Seat-based pricing is slowly dying

The logic of "charge per user" assumes humans do the work in your tool. But when an AI agent does the work, who's the seat? If one person plus an agent does what five people used to, seat-based pricing means your revenue falls as your customer gets more productive. That's backwards.

The shift is toward outcome-based and usage-based pricing: charge per resolved ticket, per generated document, per workflow run, per unit of value delivered. It's harder to forecast and harder to sell, but it aligns price with value in a world where one user can drive enormous output. Expect hybrid models — a platform fee plus consumption — to dominate.

2. AI is eating the UI

For two decades, the product was the interface — the dashboards, the forms, the carefully designed flows. Increasingly, the interface is a conversation. "Show me last quarter's churn by segment and draft an email to the at-risk accounts" replaces ten clicks across three screens.

This is existential for products whose entire value was a nicer way to click around a database. If an AI layer can do the clicking, the moat evaporates. The products that survive own something the conversation can't replicate: proprietary data, deep integrations, regulated workflows, or a network effect.

3. "A thin wrapper on an LLM" is not a business

The past two years produced thousands of products that are, honestly, a prompt and a Stripe button. They got initial traction because the underlying model was magic. They're now getting crushed, because:

• The model providers ship the same feature natively, for free.
• Anyone can rebuild the wrapper in a weekend.
• There's no data, no switching cost, no reason to stay.

The durable AI-era SaaS companies are building defensibility below the model: proprietary datasets, fine-tuned domain expertise, workflow lock-in, compliance and audit trails, and integrations that take months to replicate. The model is a component, not the product.

4. Vertical beats horizontal (again)

General-purpose tools are the most exposed to AI commoditisation. Vertical SaaS — software for dental practices, freight brokers, law firms, fitness studios — has structural advantages: domain-specific data, regulatory requirements, and workflows that a generic AI assistant doesn't understand out of the box.

We expect a wave of AI-native vertical software: deep tools for narrow industries that combine an LLM with the specific data, compliance, and integrations of that vertical. This is where a lot of the next decade's durable software businesses will be built — and it's accessible to smaller, focused teams, not just the platforms.

5. Security and compliance become product features, not afterthoughts

As SaaS handles more sensitive data and more automated decisions, buyers are scrutinising security harder. SOC 2 and ISO 27001 are moving from "enterprise nice-to-have" to "table stakes for any serious deal." The AI dimension adds new questions: where does my data go, is it used for training, can I audit the model's decisions?

Vendors who treat this as a first-class product concern — clear data handling, no-training guarantees, audit logs, regional hosting — will win deals against flashier competitors who can't answer the security questionnaire.

What this means for you

If you sell SaaS: audit your moat honestly. If your only advantage is a nice UI on top of a model anyone can call, you're exposed. Invest in proprietary data, deep integrations, and a vertical focus. Revisit your pricing — seat-based may be working against you.

If you buy SaaS: expect more usage-based pricing and plan for it. Push vendors on data handling and AI training. Favour tools that own real workflow depth over thin AI features you could replicate.

If you're building something new: the bar is higher than it was in 2015, but the opportunity in AI-native vertical software is enormous — and it doesn't require a huge team, just a deep understanding of one industry's real problems.

SaaS isn't dying. The lazy version of it is. What replaces it — software that owns data, depth, and trust rather than just a subscription — is a better business anyway.