In the wake of ChatGPT’s launch four years ago, AI-powered products have flooded the market. Coding assistants, creativity tools, customer service agents, travel concierges—all powered by ever-more-powerful AI models. And they’re growing at an unprecedented rate: on Stripe, we’ve seen AI startups reach $1 million in annualized revenue 25% faster than the previous generation of SaaS startups.

But this speed belies the difficulty of successfully monetizing AI products. AI startups are hitting revenue milestones faster amid more inherent complexity in their business models. When a product does something totally new, we need new ways to define and monetize its value. On top of that, the AI inference costs incurred completing even simple tasks can vary unpredictably. Traditional subscriptions don’t work with that kind of volatility and put companies at risk of massively under- or over-charging.

An AI pricing framework

New strategies for monetizing AI are taking shape. A recent Stripe survey found that 56% of AI company leaders reported using hybrid pricing, and 38% reported using purely usage-based pricing. Though usage-based and hybrid models are typically used by companies at different phases of growth, they’re both models that better align with the incremental value customers receive from AI products, and can more reliably cover the cost to deliver that value.

But usage-based and hybrid models are complicated. These models offer many different levers to adjust to account for an AI product’s unique features; as a result, we see a lot of variation in how they’re implemented. What do effective monetization strategies have in common? We’ve found that successful AI leaders tend to have clear answers to the following questions:

• What kind of value does my product deliver to customers, and can I charge for it in a way that covers incremental costs?
  
• How can I translate this charge metric into a pricing model that encourages customer adoption and growth, while also offering predictable revenue?
  
• What additional risks does this pricing model create, and how can I manage them?
  

We spoke with pricing experts from Anthropic, Clay, Fin, PostHog, Vercel, and Bessemer Venture Partners to understand how they’re approaching these challenges. Out of these conversations, we developed a five-step framework that helps AI companies get pricing right.

1. Determine your value metric. Conduct user research to quantify the customer benefits and outcomes your product delivers.
   
2. Set your charge metric. Choose how to bill, aligning your customers’ received value with your variable costs.
   
3. Pick your pricing model. Design a model (e.g., usage-based, hybrid) that balances revenue predictability with customer growth.
   
4. Set your guardrails. Implement controls like spending caps and alerts to manage cost risks and prevent surprise bills.
   
5. Iterate on your strategy. Treat pricing as an ongoing process, making small, frequent updates to adapt to market changes.
   

Step 1: Determine your value metric

Though it can be tempting to skip straight to how you’ll charge your customer (your charge metric), we’ve seen that successful AI monetization starts a step earlier, with defining your “value metric.” But AI products can deliver many different kinds of value. They can automate tasks, augment human performance, save money—the list goes on.

What’s the right way to describe, let alone quantify, the benefits your customers get from your product?

• First, understand customer needs. AI leaders agree that setting a good value metric starts with user conversations—lots of them. AI products tend to deliver novel benefits, which interviews, surveys, and feedback sessions can help you characterize.
  

“For us, defining value starts with asking the right questions,” explains Sydney Meheula, head of product finance at Anthropic. “We look at key areas like how many hours are businesses saving from using our technology? What are the error reduction rates they’re seeing? What risks have been mitigated? What innovation has been unlocked?”

• Make sure to focus on results. Customers will tell you a lot about how they use your product day to day—but you need to pay closer attention to the outcomes they achieve, or hope to achieve.
  

Fin is Intercom’s AI customer agent engineered for precision, speed, and reliability. When it first launched, “trust in agents’ ability to effectively answer questions was low,” says Aisling O’Reilly, who leads product and pricing for Fin. “If you asked customers to pay per conversation, and the agent didn’t do what was asked, you’d essentially be asking them to pay twice: once for the agent, and another time for the human that has to come in afterwards.”

This empathy, and a bet that they could effectively cover costs, led them to adopt their outcome-based model. While it’s not practical for every company to actually charge for outcomes (see Step 2: Set your charge metric), articulating your ideal state of alignment with customer value can help you spot new monetization opportunities.

Step 2: Set your charge metric

A charge metric is the unit of usage a business prices against to translate product consumption into revenue. Though it’s not a concept unique to AI, setting a good charge metric is uniquely challenging for AI products.

This metric should align as best as possible with your value metric, while at the same time reliably covering costs. “AI introduces variable costs that can fluctuate widely depending on the task or model used,” Jasdeep Garcha of Vercel explains. “Our approach is to separate those elements—we pass through the inference cost transparently to the customer and layer our pricing around the value we deliver. That way, our incentives stay aligned: we’re not optimizing for cheaper models, but for better outcomes.”

We’ve observed that three main categories of charge metrics emerge for AI products. Each makes a different compromise between cost alignment and value alignment: the more cost-aligned models bet that their customers will accept (or even welcome) a less straightforward link to business value, while the more value-aligned models bet that they’ll be able to accurately model costs.

• Consumption based (per API call, per LLM token): This charge metric is closely tied to infrastructure cost. This means the cost incurred per action is transparent for you but harder for your customers to link to concrete business value. This metric makes sense when your customer wants granular control over what they consume.
  
  • Example: OpenAI charges per token consumed.
    
• Workflow based (per completed task): This metric introduces more variability on cost, but it’s easier to tie to value. You can set a price for the completion of more complex tasks that consume more variable resources, such as booking a meeting or analyzing a spreadsheet. Though these may not link directly to business outcomes, they have clear value for improving business processes.
  
  • Example: Salesforce Agentforce charges per conversation.
    
• Outcome based (per successful outcome): With this metric, you only charge when your customer successfully resolves a problem using your product. This metric has the most variability on cost, but is appealing to customers because it’s easiest to tie to business outcomes.
  
  • Example: Fin charges per ticket resolved by its agent.
    

Step 3: Pick your pricing model

Once you’ve decided on a charge metric, it’s time to use that metric—or metrics, if your product has multiple distinct functions—to create a pricing strategy that facilitates growth while protecting your business from undue risk.

First, the building blocks: pricing models for AI products typically use one or both of the following kinds of fees:

• Base fee: This is a recurring platform or seat-based charge; on its own, we’d call it a simple subscription. It may include a fixed amount of usage according to your charge metrics, or include unlimited usage.
  
• Scaling fee: This is a usage-based charge, corresponding to the charge metric, that may be used alone or in combination with a recurring base fee.
  

Next, as you work through the right combination of base and scaling fees, consider how to balance the following factors:

Customer acquisition and growth

Does your pricing model encourage customers to try your product and grow their usage over time? If yours is an early-stage company with aggressive growth targets, or if you use a product-led growth model, lowering barriers to adoption can encourage potential customers to experiment.

Everett Berry, the head of GTM engineering at Clay, recalls how additional platform fees “created friction that hurt our customers’ ability to expand their usage.” Once Clay moved to a purely usage-based model, growth accelerated. “By bundling all costs into the per-credit price,” he says, “we made it much simpler for customers to adopt new features and scale with us.”

Repeatable revenue

Do customers pay enough in recurring fees that it’s possible to forecast revenue? This matters when you’re trying to make investments in hiring, research, and other long-term projects—or when you’re courting customers looking for a predictable monthly spend.

For some companies, this means a purely usage-based model isn’t a good fit. “The challenge comes at scale,” Kent Bennett, a partner at Bessemer Venture Partners, explains. “Once a customer’s bill hits $100K one month and $300K the next, [companies] can’t tolerate that variability. At that point, predictability becomes the new priority, and the pricing model must evolve into a hybrid or bundled approach to support the next phase of growth.”

Pricing model options

Considering your entire product strategy

Finally, it’s worth a moment to pause and think about how your AI offerings impact the value of your core product, and how they tie into your broader product vision. Many companies rush to maximize profits on their most exciting new features. James Hawkins, co-CEO at PostHog, takes the opposite view.

His team faced a key decision when pricing the company’s AI agent: “Is this an N+1 feature, or is it a platform component?” Hawkins and team quickly realized the agent was enhancing nearly every product. “Out of every analytics report created in PostHog, nearly 20% are now made through AI,” Hawkins explains. “This realization meant our pricing philosophy had to be to charge usage to cover our costs, not maximize profit.”

Step 4: Set your guardrails

A well-designed charge metric and a balanced pricing model can both help protect against the cost risks associated with AI products. But the possibility of unexpectedly high bills remains whenever customer usage is variable.

Additional guardrails can help keep these risks manageable. The right guardrails depend on which patterns of use are most likely to be problematic:

• Usage caps with alerts: If well-meaning customers are at risk of spending more than they intend, usage thresholds with well-timed alerts can forestall unexpected bills.
  
• Billing thresholds: To solve the same problem, you can generate invoices at certain spend milestones, and require payment of the invoice before continued use.
  
• Rate limiting: If certain tasks or queries trigger significant fluctuation in resource usage, rate limits can keep spend in check while the customer adjusts their request.
  

In all cases, proactive communication is key. Customers want to understand how their usage is translating into costs, and they want to be able to react quickly if they’re spending more than intended.

Step 5: Iterate on your strategy

Though the specifics of their pricing models may vary, successful AI companies have one thing in common: they don’t treat pricing as a solved problem.

As underlying AI model costs change, market conditions mature, and customer expectations evolve—all at record pace—you’ll need to stay nimble to ensure your pricing keeps fueling growth. Findings from a recent Stripe survey support this: the highest-growth companies were nearly three times as likely to report frequent pricing adjustments.

When and how to update pricing

The greatest danger for an AI company, according to Kent Bennett of Bessemer Ventures, is “scaling before they’ve truly accounted for their variable costs, only to realize they’re running at negative gross margins…” Though guardrails (Step 4) can help forestall danger in the short term, you can’t stop there. “The foundational business guardrail,” Bennett says, “is having the discipline for brutal, self-aware internal accounting from the earliest days.”

It’s important to keep close track of how your pricing is landing with customers, how customer behaviors are changing, and how cost pressures are evolving to maintain a positive growth trajectory. These are the issues we see most commonly, and how we’ve seen companies approach a fix:

• Customer confusion or friction: If you notice repeated questions on pricing, or underuse of certain features, your pricing may be too complex to understand. Try simplifying your packaging and clarifying documentation to foster understanding.
  
• Misaligned growth: If revenue isn’t growing with product usage or value delivered, consider revisiting your charge metric. You may also want to consider adding pricing tiers or moving to a credit-based structure.
  
• Margin pressure: If you’re facing higher-than-expected infrastructure costs, you may need to add guardrails like rate limits or caps—or even consider repricing the highest-cost activities.
  
• New product capabilities: When you add new features or expand functionality, you’ll want to account for these in your pricing model. You can redesign subscription plans, charge modularly for different features, or rebundle credits to accommodate new usage patterns.
  
• Segment-specific behavior: If you notice that different kinds of customers—consumers versus B2B, apps versus APIs—are using your product differently, you may want to introduce role-based or vertical-specific pricing plans designed with their needs in mind.
  

Strategies for rolling out pricing changes

Though pricing iteration is key to finding the right pricing model, AI leaders tell us that an experiment-driven, localized approach is necessary—the same kind of approach more typically associated with product development. At AI companies, product and pricing evolve together.

Jasdeep Garcha, who works on monetization at Vercel, has observed that “a common trap is to isolate pricing changes as large projects that involve the whole entire company.” This approach slows progress. Instead, by treating pricing changes as small, localized updates, companies can create “a high-velocity, iterative practice where you’re thinking about and improving pricing every single week.”

Everett Berry at Clay highlights another danger of large-scale rollouts: alienating longstanding customers. “When we experiment with a more fundamental pricing change, we roll it out to new users first,” he explains. “This allows us to get a good signal on its impact without breaking the workflows and cost expectations our current customers rely on.”

Our survey data shows a variety of approaches, with a majority of business leaders choosing to implement pricing changes initially to subsectors of customers.

How Stripe can help

Stripe’s usage-based billing tools let you charge and manage customers however you want—from simple recurring subscriptions to usage-based and hybrid pricing—while helping you collect and retain more revenue, automate workflows, and accept payments globally. And for the most complex needs, Metronome, a Stripe product, gives you the tools to handle sophisticated usage-first models and sales-led scenarios.

With Stripe, you can:

• Offer flexible pricing: Launch quickly with any usage-based and hybrid pricing models—including flat-fee plus overage, credits, and more. Support is built in for coupons, free trials, prorations, and add-ons. As you scale, Metronome enables multidimensional billing and lets you customize rate cards by customer segment, update pricing instantly, and manage complex negotiated contracts.
  
• Monitor real-time trends: Access real‑time analytics for usage and subscriptions to get a complete view of business performance. Quickly spot wins and improvement areas, and benchmark against similar businesses on Stripe. Metronome adds event-level usage visibility for cross-functional teams to track detailed consumption, identify spending patterns, and streamline revenue recognition.
  
• Experiment and iterate on pricing: Respond to user demand faster with no-code tools to adjust usage-based rates, manage pricing cohorts, and inform pricing decisions with granular usage and spend analytics.
  
• Align pricing to customer value: Meter and charge by the usage dimensions that deliver the most impact, and define pricing in ways that directly reflect how customers gain value.
  
• Increase revenue and reduce churn: Improve revenue capture and reduce involuntary churn with AI-powered Smart Retries and recovery workflow automations. Stripe recovery tools helped users recover over $6.5 billion in revenue in 2024.
  
• Boost efficiency: Use additional Stripe solutions for tax, revenue reporting, and data to consolidate multiple revenue systems into one. Easily integrate with third-party software.
  

To learn more about how Stripe can help your company price for growth, contact our team.