Why Value-Based Pricing of AI SaaS Product Failed (And What Actually Works)
A B2B SaaS Story of How Perfect Theory Meets Enterprise Reality
👋 Hi, it’s Antti Latva-Koivisto with a new article on pricing of AI-based B2B SaaS products. If you like it, forward it to your colleague!
They were certain they’d nailed it.
The product team was ready to launch their AI-powered solution with what seemed like the perfect pricing model. AI would automate analysis and decisions, and enterprise customers would pay €2 only for successful outcomes. No success, no charge.
“We had what we thought was a good idea,” the Chief Product Officer told me. “The pricing would be based directly on the value received by the customer.”
The team was confident. It was elegant. It was fair. It aligned incentives perfectly.
Or so they thought.
Why Value-Based Pricing Seems Inevitable for AI
If you’ve been anywhere near B2B SaaS in the last few years, you’ve heard the gospel of value-based pricing. Experts have consistently recommended this approach: “Don’t charge for seats! Don’t charge for features! Charge for value!” I’ve expounded it myself too.
Value-based pricing seems especially perfect for AI products. As Kyle Poyar explained in his October 2024 newsletter article on AI monetization, “We’re moving away from charging for access to software and toward a model of charging for the work delivered.”
AI fundamentally changes what’s being sold. Instead of humans logging into software, algorithms perform work directly with perfectly measurable outputs: tasks completed, conversations held, predictions made. This creates a natural opportunity to align pricing with value created rather than seats purchased – a pricing holy grail that was previously elusive in traditional software.
The evidence is compelling:
Intercom pioneered paying per resolution with their AI customer support agent at $0.99 per successful resolution
Zendesk followed with a similar model of charging per successful autonomous resolution
Salesforce went all-in on “Agentforce” at $2 per conversation
11x charges per task completed by AI sales development reps
According to Poyar’s research, this value-based approach allows companies to capture 20-25% of the economic value they deliver, compared to just 10-15% with traditional subscription models (see image below).
Our CPO’s team wasn’t making a rookie mistake. They were following the playbook written by the industry’s forward thinkers.
Reality Check
“Our pricing was done wrong, to put it bluntly.”
There’s no sugar-coating it. What should have been a triumphant launch instead led to months of struggle.
“It wasn’t a successful launch. It took a long time until we reached a commercially good place with it,” the CPO admitted, his voice carrying the weight of those difficult months.
The team had built exactly what pricing theory recommended, but the market rejected it with startling clarity.
What went wrong? The CPO didn’t mince words:
“In a way, it was of course a very elegant and nice idea. But it was quite complex. And then it wasn’t predictable.”
Two simple problems: complexity and unpredictability. But they were enough to derail the entire strategy.
The Counterintuitive Truth
What happened next defies the tenets of value-based pricing.
“When we later changed our pricing model to a pure subscription fee, which wasn’t in any way volume-based, we started selling much more.”
The revelation is jarring. After months struggling with their theoretically perfect value-based model, they switched to a simple flat subscription – the very approach experts were saying was outdated for AI products – and suddenly sales took off. I was surprised to hear this.
“It sounds foolish. Why wouldn’t a customer want to pay only for the value produced?” the CPO reflected. “But ultimately, enterprise customers valued predictability more. They wanted to know how much this would cost them.”
This revelation aligns perfectly with what Kyle Poyar documented in his January 2025 article, where he noted a critical emerging pattern: “As AI agent businesses continue shifting away from seat-based pricing and toward charging for units of work delivered, they tend to encounter a nagging objection from customers: your pricing is too complicated and too hard for me to predict.”
Enterprise Procurement: Why Finance Says “No”
When I pressed deeper on why predictability matters so much to enterprises, the CPO revealed insights into the organizational dynamics that value-based pricing models overlook:
“Conversations about purchases tend to be more difficult with the CFO when the price isn’t fixed,” he explained. “If you say this will cost something between €30,000 and €60,000, finance immediately asks: ‘Is €60,000 the absolute maximum?’ ‘Well, no.’ ‘Then what is?’ It just becomes a more difficult conversation.”
The CPO highlighted a critical mismatch between how costs and benefits materialize: “The savings come from working hours. The savings don’t directly materialize into anything, you need to take additional actions. You need to either reduce staff or move them to other tasks.”
This is why value-based pricing creates friction: “When you have a directly variable cost [value-based pricing tied to outcomes] that’s supposed to save slowly-changing costs [labor], the equation in the decision-making situation is not so simple.”
Practitioner Insights Ahead of Industry Best Practices
Poyar’s thinking has evolved on this issue. In his October 2024 article, he acknowledged potential challenges: “Traditional enterprise procurement departments are (probably) 0% prepared for an unpredictable bill—even if that bill is better aligned with ROI.” But this was a brief aside, not a central concern.
By January 2025, however, this had emerged as a primary objection to value-based AI pricing – exactly what our CPO had discovered through painful experience many months earlier.
The industry's best practices are catching up to what practitioners have already learned the hard way: enterprise buying decisions aren’t just about economic value – they’re about navigating buyers’ preferences and organizational complexity.
How to Balance Value and Predictability?
So how did our CPO's company turn things around?
They simplified, dramatically. Their “pure subscription fee” removed all the complexity that had been scaring away potential enterprise customers. No usage metrics, no outcome calculations, just a straightforward monthly cost.
This runs counter to much of the current advice about AI pricing, but the results speak for themselves.
That said, I think the industry shouldn’t give up. Value-based pricing can be advantageous both for the software vendor and the customer. There’s potential both for a lower total cost of ownership (TCO) for the customer and greater pricing power for the vendor.
Some companies are indeed finding creative ways to address the predictability problem while still capturing more value than what’s available with a pure subscription-based pricing. According to Poyar’s latest research, several approaches are emerging.
The AI “FTE” Model
One solution Poyar highlights is positioning AI as a “Full-Time Employee (FTE) equivalent.” Companies such as 11x map the typical output of a human employee (for instance, an SDR might research X accounts and send Y emails), package it as a SKU, and price it at 20-35% of what hiring a person would cost.
As Poyar explains: “It translates complex pricing into something that feels predictable and value-based. Buying AI credits for ‘tasks’ completed by an AI SDR? That feels like a lot of math. Buying the equivalent output of a human SDR? That’s much easier to wrap your mind around.”
Skill-Based Pricing Tiers
Another approach Poyar identifies is moving from feature-based to skill-based pricing tiers. If the customer is hiring AI to do a job, software vendors must un-gate access to whatever features are needed to get the job done. More access to features leads to more jobs done which leads to more money, particularly if the pricing is based on units of work. Customers are willing to pay for higher skilled work:
Good: “Do the job cheaply” (think intern level)
Better: “Do the job well and for a fair price” (college grad level)
Best: “Do the job quickly and perfectly” (experienced hire level)
Instead of gating features, companies charge more for higher skill levels, faster outputs, greater accuracy, or human verification.
Creative Enterprise Accommodations
Poyar also suggests ways to make usage-based pricing more palatable to enterprise procurement:
Annual draw-downs: Letting customers flexibly use their pre-purchased usage over 12 months. If they use the product faster than expected, they have time to plan and budget before renewal.
Grace periods: Customers can either upgrade their contract at a higher commit or pay for the one-time flex spend after the grace period.
Roll-over options: Allowing unused credits to transfer to the next contract period
The Discovery Gap: What To Do Earlier
This pricing saga reveals a blind spot in product discovery. While functional discovery likely happened – understanding what outcomes the AI needed to deliver – organisational dimensions were missed.
In B2B SaaS, the purchase decision-maker is usually not the stakeholder who benefits from the product. Effective product discovery needs to extend beyond functional needs to include organizational buying dynamics:
Procurement process mapping: Understanding approval steps and thresholds
Budget cycle understanding: Learning how and when budgets are set
Finance stakeholder discovery: Directly engaging financial decision-makers
As one founder I interviewed explained: “In B2B, there are several parties that have an interest in the product. It is not necessarily the interest of only one type of potential customer, like the CMO, but also the CTO, for example. A more comprehensive understanding of individual customers helps find a Go-to-Market model to penetrate the target market segment.”
Good product discovery must uncover both dimensions:
What problem does the product solve and how? (Problem-Solution Fit)
How will customers actually purchase it? (Go-to-Market Fit)
Even the perfect solution will fail if it conflicts with how enterprises make purchasing decisions. A product must fit not just the functional customer problem but also align with organizational realities around budgeting, approval processes, and how value is measured and realised.
Finding failure only after launch is extraordinarily expensive. Investment in comprehensive discovery (a few weeks or months of work by 1-3 people) pales in comparison to the cost of building, launching, and then having to pivot a product because of either misaligned Problem-Solution Fit or Go-to-Market Fit. This company spent months waiting for revenue to materialise, while burning marketing, sales and engineering resources. The opportunity cost was even higher, and could have been avoided with better product discovery.
What This Means for AI Pricing Strategy in Enterprise SaaS
This experience offers several critical insights:
Predictability trumps perfect alignment. Enterprise customers will choose a slightly less efficient pricing model if it makes budgeting easier.
Consider your customers’ internal processes. Enterprise purchase approvals often require fixed budgets set months or a year in advance.
CFO psychology matters. Costs are carefully budgeted, but revenue is less controllable. CFOs hate unpredictable expenses, even when theoretically tied to value.
Organizational procurement realities create friction. The stakeholder who gets the benefit (e.g. in business operations) is often not the same person who approves the budget (finance). This creates misaligned incentives in the purchasing decision.
Value realisation has a time lag. When the CPO said that “savings don't directly materialize into anything,” he’s highlighting that the benefits of many AI-based products require organizational changes to be realised – changes that may take months or even years, especially in markets with strict employment regulations.
Test your pricing assumptions. Don’t assume any model is optimal without validation, no matter how logically sound it seems.
Simplicity sells. Every layer of complexity in your pricing creates friction in the sales process.
Be willing to pivot. Our CPO’s company saved their product by recognizing reality and adapting their model.
The AI pricing revolution isn’t over. But as it evolves, the winners will be those who balance the appeal of value-based pricing with the practical reality that for many enterprises, predictability still trumps perfect value alignment.
The art is finding the model that your customers will actually say yes to.
Did you get value out of this article? Share it with your colleagues!
This article draws from my in-depth discussions with B2B SaaS product executives and founders. While names and other identifiable details have been anonymized to protect confidentiality, quotes are drawn directly from our conversations.