I’ve spent more than 20 years helping companies price and package subscription offerings, and I don’t think I’ve ever seen a shift as destabilizing as the one created by AI. 

For years, SaaS pricing (and growth) forecasts were built on this simple logic: If your customers made more money, they hired more people, and needed more seats. Revenue growth at your customer meant revenue growth for you.

But that equation is breaking down. Across industries, companies are growing revenue without growing headcount. According to Gartner, enterprise software buyers increased software spend by nearly 13% in 2025, while employee counts stayed almost flat. Revenue growth doesn’t lead to headcount growth the way it used to.

AI is also doing more of the work: drafting content, analyzing data, qualifying leads, automating workflows. But that’s coming at a heavy variable cost for vendors — something that was never part of the classic SaaS margin story.

By 2026, most SaaS products will be AI-enabled, and many will rely on AI to deliver the primary value customers pay for. The problem: pricing models aren’t keeping up. They’re mostly still seat-based, feature-tiered, and not aligned with actual outcomes (what customers generally want). 

If nothing changes, SaaS companies will see gross margins deteriorate and growth slow down. Many are rushing to raise prices because of margin compression. But if you do too much too soon, customers might not even bother with trying your AI. What’s more, changing pricing models is a heavy undertaking, which makes this a hard needle to thread for everyone.

Yet, I think there’s an opportunity here to take a step back and rethink core pricing questions — prioritizing the customer’s perceived value, your costs, and how you can equitably and profitably price your products so they get value and you get margin.

Here are 7 guidelines, listed in priority, based on my own work as well as from observing how mid-market and enterprise companies have already been experimenting with real-world pricing models. 

Let me know what you think!

Before you start fiddling with pricing, it’s important to understand your own changing costs, ideally at a unit economics level.

Legacy SaaS is high gross margin, with low variable costs. AI-first SaaS is not — at least not yet. Inference compute (the computational power needed for an AI model to use its learned knowledge to make predictions on new, unseen data) is expensive. Data storage is expensive. Fine-tuning models is expensive. And vendors who price as if AI were “free” (in the traditional SaaS sense) risk collapsing their unit economics.

This is why Adobe has raised prices and explicitly stated that the cost increases tie to the costs of generative AI. They learned early (and publicly) that without guardrails, customers might generate thousands of images per day and expect it to be included. 

It might be painful to meter your offerings or talk about your own costs, but by being candid about why your pricing is changing, you might actually build, not erode, trust.

For mid-market companies, the takeaway is simple: Price in a way that protects gross margin while aligning with perceived value. AI isn’t like traditional SaaS, where incremental usage was essentially “free.” If you need to test how your AI functionality will be valued by customers, do small, bounded tests before launching inclusive pricing broadly. And if you can’t afford to do these tests, and price against your own costs, communicate clearly and be prepared for customer pushback and slower adoption.

Don’t leave the task of explaining your pricing changes and experiments to your customer success team. This is a strategic GTM issue that will impact your relationships with your best customers. Bring in product marketing to help customers understand where you’re going with your product and why.

One of the most consistent objections I hear from SaaS buyers right now is the fear of an unpredictable bill. Volatility feels risky.

There are a lot of ways to provide greater predictability, even as you charge against your own costs. For instance, you could build dashboards to make it easy for customers to see what they’re spending, with spending caps and automatic throttling that they can set.

Long before AI, Snowflake mastered this approach by giving customers exceptional visibility into their costs to run their data warehouse. As the company has moved into AI workloads, they’ve kept the same philosophy: transparency creates comfort; comfort drives usage; usage drives value and retention.

As you introduce your own AI features, the same rule applies. Don’t surprise customers with unpredictable usage charges. Instead, invite them into a controlled environment. It’s the same principle that governs great subscription businesses: members stay when they feel in control.

Pure usage-based pricing makes buyers anxious, and you want to make it easy for the buyer to buy. But a simple “all you can eat” subscription creates margin risk for vendors — especially as the operational costs of using a trained AI model (i.e., the inference costs) continue to fluctuate.

One solution: share the risk with your customer. 

OpenAI’s ChatGPT Team and Enterprise offerings are good examples of this compromise. The Team plan is $25–$30 per user per month, but the real scalability happens through metered API consumption. 

HubSpot is moving in a similar direction. Their 2025 announcements emphasize usage-linked expansion for AI features such as email sends, AI-generated customer insights, or enriched CRM records. Customers pay a base CRM platform fee, but incremental costs follow usage patterns.

This model reflects the core principles of my work in the Membership Economy:

Customers don’t care about access or tokens or seats; they care about outcomes. 

This isn’t just an AI issue; it’s a more general pricing issue. It’s easier to price against our effort and costs, because it’s harder to measure the value to the customer. 

But if you solve an important problem for the customer, and price against that problem’s solution, customers will be less price sensitive, and you’ll be more generously compensated for the value you provide.

Customers do understand “cost plus” and, especially during this transitional time with AI, they may be willing to pay for use. But when it’s time to scale, they’ll want evidence of impact.

Apollo.io now uses AI to score leads. The more contacts you upload, the more value you get. You’re really paying for getting better leads — not just for using the tool — so they charge based on credits tied to each enriched profile (i.e., outcome-based pricing). 

However, pricing around outcomes only works when the outcome is clear and easy to measure. If it isn’t, this kind of pricing can frustrate customers. You really can only use outcome-based pricing when the outcome is straightforward. 

Ask yourself:

It’s not enough for you to price for the potential value you might be creating — the customer has to agree. People renew when they feel they’re treated fairly. Pricing should reinforce that trust, not erode it.

AI creates tons of activity logs. “Tokens used.” “API calls made.” “Credits consumed.” Those metrics mean almost nothing to customers. But if you can translate those activity logs into meaningful business impact, you’re onto something. Think about hours saved, costs avoided, output improved, backlog reduced, or revenue influenced. These represent value worth paying for.

Figma, which introduced “Figma AI” this year, did just this. They show users not just how many AI generations they ran, but what those generations accomplished, such as components created or drafts completed. 

Reframe AI’s work in terms the end user recognizes as valuable. Tying usage to impact strengthens your renewal story. A product that repeatedly reminds its customers of its value doesn’t need to be discounted to stay relevant.

At the Vegas buffet, some people pile on the shrimp and prime rib, while others stick with pasta and bread. But they pay the same, and the restaurants know they’re OK with it because they understand the usage patterns.

AI usage is similarly varied. Projects, campaigns, onboarding periods, analysis windows — all of them create short bursts of heavy consumption. And some customers need more than others at certain phases. 

A rigid plan penalizes natural cycles in a way that makes cancellation more appealing than renewal. And pricing purely on usage puts a lot of risk on the shoulders of your customers, creating complexity that makes them nervous about leaning in.

As AI becomes woven into the core of every serious SaaS offering, we can’t rely on the pricing structures that worked when value roughly correlated with the number of people logging in. Your customers’ businesses are growing faster than their headcount, their reliance on automation is increasing, and your own costs now vary month to month based on compute and data activity. Seat-based pricing simply cannot absorb that complexity — and it certainly can’t communicate value.

The path forward is not to chase a single “right” model, but to experiment intentionally with pricing that reflects how your product actually creates value today. Start small. Identify one or two measurable outcomes customers care about — leads enriched, hours saved, documents generated — and pilot models that align revenue to those outcomes. Introduce a hybrid plan that protects predictable recurring revenue while giving customers a clear line of sight into how usage affects spend. Build transparency into every step: dashboards, caps, explanations, and cost drivers they can understand without a PhD in Machine Learning.

Most importantly, talk to your customers. Understand how AI is reshaping their workflows, where they’re replacing headcount with automation, and what business results they now attribute to your product. Use that insight not only to tune pricing but to strengthen the relationship and reinforce what they’re really “joining” when they do business with you — a partnership where incentives are aligned and value is visible.

And remember: You don’t need to overhaul everything at once. But you do need to begin shifting from pricing access to pricing outcomes. The companies that make this transition thoughtfully will build deeper loyalty, clearer differentiation, and healthier economics — even as AI continues to change what software can do and how customers expect it to be priced. 

In a world where teams are leaner and results matter more, aligning your revenue model with the value you deliver isn’t just the right strategy. It’s the only one that’s built to last.

Become a Topline insider by joining our Slack channel.

We want to hear your feedback! Let us know your thoughts.