SaaS Pricing in the Age of AI

The Pricing Upheaval of the AI Era

You've built a SaaS product. How do you price it? Do you go with a flat monthly subscription, charge by usage, or give it away free and switch to paid past a certain point? This question isn't just about setting a number—it's a choice that determines the entire business structure of your product.

The problem is that we're in the middle of an upheaval in SaaS pricing. As AI moves in, the golden formula of SaaS—the seat-based subscription—is being shaken. As David George of a16z diagnoses it, the first cost customers try to cut when they adopt AI is headcount. And seat-based pricing is precisely the target.

Here we reframe the nine models laid out in February 2026 by Schematic, a software monetization platform, from the perspective of a founder. Think of it as a yardstick for judging which combination fits your own product.

An Overview of the Nine Pricing Models

Look at each model closely and it becomes clear what fits where.

1. Flat Subscription: The Simplest, but Growth Stalls

Unlimited use for, say, $300 a month. It's the simplest model. Easy to sell, easy for customers to understand, easy to forecast revenue.

The trouble is that revenue doesn't grow even as usage climbs. Light users and heavy users pay the same. That's a win for the heavy user, but from the company's side, you can't generate additional revenue from the very customers extracting the most value.

Most companies lay down a flat subscription as the base and then expand on top of it with usage charges or add-ons. Going with a flat subscription alone is increasingly rare.

2. Seat-Based: The Golden Formula of SaaS—and Now in Crisis

For the past 15 years this was the SaaS standard. Twenty-five dollars per user per month. More users, more revenue. Fewer users, less revenue. Simple and predictable.

Collaboration tools (Slack, Notion), design tools (Figma), and CRMs (Salesforce) all grew on a seat-based model.

But in the AI era this model is being challenged head-on. David George puts it precisely: "The first cost customers try to cut when they adopt AI is headcount. And seat-based pricing is precisely the target."

If a tool that 100 employees used can now do the same work with 50 employees plus AI agents, the seat count is cut in half. Same customer, but revenue halved. SaaS companies that rely on seats alone are likely to be the first casualties of the AI era.

3. Usage-Based: A Likely Standard for the AI Era

Number of API calls, events processed, compute time, data volume. You charge only for what the customer actually uses.

AWS, Twilio, and Stripe built enormous businesses on this model. Whether users go up or down, revenue rises as activity rises. Even when AI agents shrink the seat count, the API calls those agents generate go up. It's a natural fit for the AI era.

The drawbacks are just as clear, though. From the customer's side, the monthly bill is unpredictable. The anxiety of "What's it going to be this month?" can block adoption. That's why usage-based pricing is usually designed as a baseline plus overage usage.

4. Credit-Based: The Best Fit for AI Products

This is fast becoming the standard for AI products.

You buy credits up front, and credits are deducted each time you run a task. Light tasks deduct a little, heavy tasks deduct a lot. ChatGPT's message credits, Midjourney's GPU time, the Claude API's tokens—they all share the same structure.

Why are credits a good fit for AI? Because the cost varies wildly from task to task. A simple text request and a video generation can differ in cost by a factor of 100. Charge purely by usage and the bill swings too much. With credits, you get predictability—"50,000 credits = $50 a month"—while still deducting in proportion to the real cost of each task.

In Korea, too, if you're building AI-based SaaS, the credit model is worth considering first. Users can control their own consumption, and the company can manage volatile infrastructure costs.

5. Tiered Plans: The Most Familiar Packaging

Starter, Pro, Enterprise. Three-step packaging is the standard.

The lowest tier protects existing revenue. The middle tier absorbs expansion. The top tier is the space for negotiation and custom terms.

There's a reason three steps became the standard. Customers can compare them at a glance, sales reps can explain them easily, and the price differences are clear. Four or more tiers actually induce choice paralysis.

That said, this model is rarely used on its own. It's usually designed as tiers + usage limits + add-ons.

6. Add-Ons: How to Keep the Core Product Light

Audit logs, advanced permission management, premium integrations, higher API limits. These are things not every customer needs.

Break them out as add-ons and the core product stays lean. Pricing stays simple too. "The base is $29 a month; if you need audit logs, that's +$10." Customers buy only what they need, and the company generates incremental revenue.

The real value of add-ons lies in lowering the burden of shipping new features. You don't have to redesign your entire pricing structure every time you build something new—you just add it as a new add-on.

7. Freemium: The Most Powerful—and the Most Dangerous

You start free and convert to paid once a certain limit is crossed. Notion, Slack, and Dropbox grew on this model.

The upsides are clear. The barrier to entry is zero. It spreads by word of mouth. You invite users to convert to paid only after they've grown comfortable with the product.

The downsides are just as clear. Set the limits wrong and only free users pile up while revenue stays flat. Infrastructure costs rise because of free users, and if conversion to paid doesn't follow, losses accumulate.

The crux is where you set the limit. Too tight and users leave before they ever arrive. Too loose and they use it free forever. The point that naturally nudges users to convert right after they feel the product's true value—that's the heart of freemium design.

8. Marketplace Fees: The Standard for Platform Businesses

This model takes a fee when a transaction occurs. Platforms like Uber, Airbnb, and Coupang are built this way.

Fifteen percent per transaction, or a flat fee, or a paid listing charge. You don't force users into a subscription, yet revenue follows as transaction activity grows.

There's a difference from SaaS. A marketplace has to build a two-sided market. It only works once you've gathered both suppliers and buyers. It's far harder to stand up at the start than SaaS, and revenue can be near zero until you reach critical mass.

9. Performance-Based: A New Possibility for the AI Era

The most interesting model, and the hardest. You're not selling access—you're selling outcomes.

An AI sales tool charges $40 per lead acquired. An AI fraud-detection tool charges a fee per fraudulent transaction blocked. An AI customer-support bot bills per inquiry resolved.

From the customer's side, it's the most attractive model. No results, no payment. From the company's side, you can tie the value of your product to revenue as directly as possible.

The problem is measurement. How do you define the "outcome"? How do you prove that outcome happened because of your product? If measurement is imprecise or disputes arise, the whole model collapses.

As AI automation products proliferate, this model is likely to grow, because AI works in domains where clear outcomes—the number of cases processed, the time saved—are easy to measure.

The Rise of Hybrid Models

Almost no company picks just one of the nine above. Most combine two or three.

There are some common combination patterns.

Subscription + usage overage. Unlimited up to a usage cap inside the base plan, then per-usage charges once you exceed the cap. Claude and ChatGPT Enterprise use this structure. It captures predictability and scalability at the same time.

Subscription + credits. A flat subscription provides the core features, and AI tasks are deducted in credits. An increasingly common pattern in SaaS with a side of AI.

Tiers + add-ons. You lay down three tiers and sell extra features as add-ons on each tier. The shape of standard B2B SaaS.

Self-serve + sales negotiation. Small customers sign up themselves by credit card; large customers get custom contracts through a sales team's negotiation. The same product sold through two different channels.

The crux is a structure where subscription creates predictable revenue and usage/credits capture expansion revenue. You need both.

How to Choose a Model

The answer to this question lies in "how your customers use your product." Look at three patterns.

If customer usage is steady, go subscription or seat-based. Tools that users rely on similarly every day. Email clients, CRMs, and collaboration tools fall here.

If customer usage is variable, go usage- or credit-based. Tools that might be used 100 times in a month or 10,000 times. AI APIs, infrastructure tools, and data-processing tools fall here.

If usage patterns differ greatly from customer to customer, go tiers + add-ons. Cases where a small team and a large enterprise use the same product in different ways. Most B2B SaaS falls here.

Why You Need to Rethink Your Pricing Model

Until now, SaaS pricing was a decision that, once made, stuck around for a long time. Pick seat-based once and it ran unchanged for five years.

The AI era is different. Seats shrink. The load AI generates isn't proportional to the number of users. The way automation produces outcomes differs from human labor. The assumptions of existing pricing models are breaking down.

To quote David George's diagnosis again, he said there are two paths: lift your revenue growth rate with AI-native products, or redesign your cost structure to lift your operating margin. On both paths, redesigning the pricing model is central.

If you're running a SaaS today, you need to check whether your pricing model still works in the AI era. If you lean on seats alone, you should add usage or credits. If you're going with a flat subscription only, you should build a structure that captures additional value from heavy users. The era when one model was enough is over, and an era where hybrid design is the standard has arrived.

If you're starting a new SaaS, it's best to design for a hybrid from the outset. Subscription for the baseline, usage or credits for expansion. Start with both of these axes in hand and you can respond flexibly to shifts in the market.

Price is not just a number. It's the business model itself—how you convert the value your product creates into revenue. In the AI era, the companies redesigning that model are the ones who will win the next five years.