A Behavioral Science-Driven Framework For Pricing AI Products

Written By Robyn Risso

We’ve recapped the key insights from Kristen Berman, CEO and founder of Irrational Labs, behavioral scientist, and pricing expert.

How do you get people to pay for AI? That was the central question Kristen Berman, CEO and co-founder of Irrational Labs, tackled in a recent Enrich event on AI pricing and monetization. Drawing on behavioral science research and real-world experiments, Kristen delivered a sharp, practical framework that challenges conventional wisdom about willingness to pay.

Here are the key insights and what they mean for your pricing strategy.

1 - Value Is Not Inherent, It's Created

Kristen opened with a deceptively simple exercise: identifying the most expensive purse from a lineup. Nobody got it right. The lesson? Value isn't determined by the physical product, but shaped by branding, description, context, and market signals.

The same logic applies to software. There's no fixed willingness to pay for your product. It's constructed by you, your market positioning, and the context you create around it.

Takeaway: Don't assume your customers have a fixed price ceiling. The anchors and signals you create around your product directly shape what they're willing to pay.

2 - People Pay For "Right Now" - Not a Future Benefit

Behavioral science has a clear finding: humans are present-biased. Telling a customer that your AI will make them 30% more productive over the next year is the equivalent of telling someone to brush their teeth to avoid cavities in old age. They know it's true. They still won't act.

The motivating benefits that drive willingness to pay are immediate and emotional. For example, Verizon locks people into 3-year contracts with a free phone today. The long-term value of the Verizon plan is not how customers are making the buying decision.

Takeaway: Anchor your pricing and upgrade prompts to immediate, tangible value, not long-term ROI. Ask yourself, what will this do for someone in the next hour?

3 - "AI" as a Label Doesn't Move the Needle

Irrational Labs ran a controlled study: they showed nearly 1,000 participants landing pages for real products, half with AI prominently featured, half without. The results were striking. Labeling tools as AI had no significant effect on willingness to pay, no significant effect on trust, and no significant effect on perceived performance improvement.

In the behavioral science motivation matrix, "AI" falls in the bottom-left: vague, future-oriented, not emotionally compelling. It doesn't make someone act today. 

Takeaway: Stop leading with AI as a selling point. Lead with the outcome. People don't pay for AI, rather they pay for faster language learning, better meeting notes, or an hour back in their day.

4 - Make It Unusable Without Paying

Some of the most effective freemium-to-paid conversion strategies aren't about persuasion at all, but rather they're about friction. Some familiar examples? Canva, Descript, Lovable, and Typeform all use watermarks on exported assets. Zoom imposes a 40-minute cap. Slack deletes message history after 90 days.

These mechanisms are effective because they create a social or functional pressure to upgrade. Users don't blame the company, they blame themselves. This can drive an actual willingness to pay.

Takeaway: Identify the moment your product will cause friction if it’s continued to be used for free. Design your paywall around that moment, not around feature lists.

5 - Usage-Based Pricing Makes People Use Less

Usage-based pricing is increasingly common in AI, and for good reason given the cost structure. But it has a serious behavioral drawback. When people are charged per token, per slide, or per task, they pay attention to each unit of consumption. That attention causes them to use less.

Kristen used a memorable analogy. If you charged someone per bite of pizza, they'd eat fewer bites. Charge for the whole slice, and they eat the whole slice.

The solution isn't to abandon usage-based pricing, but change the messaging to change the behavior. Instead of "2,000 tokens," give users "10 presentation decks." Package usage in terms of workflows and outcomes, not technical units. This preserves the certainty premium (people pay more for predictable costs) without triggering anxious rationing behavior.

Takeaway: If you use usage-based pricing, translate credits into meaningful outputs. Users should understand what they're getting in human terms, not token counts.

6 - People Pay a Premium for Certainty

One of the most consistent findings in behavioral economics is that people will pay more to avoid uncertainty, even when the uncertain outcome might be better on average. Applied to pricing, this means that unpredictable bills create anxiety, and anxiety suppresses engagement and conversions.

Snowflake learned this the hard way. Early customers were charged after-the-fact for data usage and hated it. To reduce friction, Snowflake added upfront estimates and usage modeling to help customers plan. The lesson is that any company moving to usage-based pricing must also invest in helping customers predict and control their spend.

This is why Claude and ChatGPT's admin controls, which let enterprise admins allocate specific credit budgets per team per month, are a meaningful differentiator. They give buyers control, which creates confidence.

Takeaway: Build tools that help customers forecast what they'll spend. Certainty is a product feature, and it’s one that people will pay a premium for.

7 - Seat-Based Pricing Is Under Pressure

The traditional SaaS model (charge per seat) is increasingly strained in the AI era. Kristen identified three pressures. First, if some employees don't use the AI features, companies feel they're paying unfairly. 

Second, AI usage is inherently spiky, making a uniform seat price feel wrong. 

Third, and most disruptively, if AI agents replace human seats, pricing on human seats becomes a shrinking revenue model.

Alternatives are emerging. Figma's tiered seat model (viewer, collaborator, full seat) lets companies pay for actual engagement levels. Lovable is experimenting with a unified credit pool shared across an organization rather than tied to individuals.

Takeaway: If you're building B2B AI products, design your pricing model to be flexible. Build pricing re-evaluation milestones into contracts, and start thinking about how to charge when agents, not humans, are the primary users.

8 - Frame What You're Selling as an Outcome, Not a Process

Kristen drew a contrast between two types of AI features, process features ("AI summarizes your sales call") and outcome features ("summary auto-syncs to Salesforce"). The outcome version costs less to deliver as it's often just an integration wrapper, yet people are willing to pay more for it.

This seems paradoxical, but it reflects how humans make decisions. We pay for results, not effort. An AI that does something noticeable and concrete feels more valuable than one that labors quietly in the background.

Duolingo charges significantly more for its AI-enhanced "Max" tier, not because it's selling AI, but because it's selling better language learning. Alpha School charges premium tuition not because they use AI, but because they promise better outcomes for children.

Kristen explained, "People aren't paying for AI. They're paying for an hour more free time, or to learn a language better, or to give their kid a great education."

Takeaway: Anchor your pricing to the outcome your product delivers, not the AI tool powering it. The more concrete and personal the outcome, the higher the willingness to pay.

9 - The Biggest Pricing Mistake: Forgetting Human Psychology

In the AI era, companies are increasingly pricing based purely on cost structure, asking, “What does it cost us to serve this feature?” 

But Kristen's core argument is that this approach leaves significant revenue on the table.

The behavioral drivers of willingness to pay - reciprocity, fairness, status, social norms, concreteness, and prior reference points - didn't disappear because AI is expensive to run. Companies that ignore these levers in favor of pure cost-based pricing miss the opportunity to charge what their product is actually worth to users.

Takeaway: Don't let infrastructure costs be the only input to your pricing decisions. Audit your product for the behavioral signals that drive willingness to pay: what makes users feel it's fair? What creates reciprocity? What signals status?

10 - Pricing Will Become a Competitive Differentiator

Kristen's parting prediction? In a world where AI features are increasingly commoditized, pricing itself will become a differentiator. The companies that offer predictable, fair, clearly-structured pricing will command a premium. This is especially true in enterprise, where procurement teams are trying to model multi-year costs and can't afford ambiguity.

Southwest Airlines built a loyalty advantage not through service quality but through "Bags Fly Free." It was a pricing decision that created a perception of fairness and transparency in an industry full of hidden fees. The same opportunity exists in AI.

Takeaway: Treat your pricing model as a product. Invest in making it clear, fair, and predictable, and watch it become a reason customers choose you over competitors with better features but more confusing bills.

Final Thought

The through-line of Kristen's talk is both humbling and empowering: human decision-making is irrational in predictable ways. Pricing isn't just math and margin, it's psychology. The companies that study how their customers actually make decisions, and design their pricing models accordingly, will consistently outperform those that rely on traditional models and pure cost-plus logic.

In the AI era, where cost structures are volatile and customer expectations are still being formed, this behavioral science lens may be the most important pricing tool available.

Interested to hear more from Kristen and her team?

Robyn Risso
Next

The Future of Work for Developers in the Age of AI