Pricing is a Growth Lever, Not a Number

What moving from user-based to volume-based pricing taught me about growth in our AI startup.

User-based pricing quietly break business models for AI-native products.

It's not because pricing users is inherently bad, but it's because it prices the wrong unit of progress for many AI companies.

We learned this the hard way while making a pricing transition at NEXT AI, where we moved from a per-user license model to a volume-based model for all our new and existing customers. What started as a simple pricing model change ended up completely reshaping adoption, customer behavior, and how we were able to deliver value.

This post is a reflection of what happens when pricing stops being a line item and starts behaving like a growth lever – and for us, one of the best decisions we made to date.

The moment we realized user licenses were the wrong unit

At its core, the product we're building isn't a collaboration tool or AI productivity app. It's a customer insight engine that ingests messy qualitative feedback, processes it with AI, and turns it into insights teams can act on.

The value we deliver increases the more people can benefit from it. And yet, we were pricing access.

On paper, things looked fine. Customers were genuinely excited by what we built, demos landed very well, and our vision resonated with senior stakeholders. However, growth, especially in existing accounts, was stalling.

As we started digging into the problem, a pattern was emerging. The moment pricing came up in conversations, energy shifted. People weren't skeptical about our value proposition, but it shifted from ambition to an accounting conversation.

• “Who really needs access?”
• “Can we start with fewer users?”
• “Do stakeholders actually need to log in?”
• “We have 40 PMs, but maybe we can share some licenses across them?"

The truth is, we weren’t fighting competition. We were fighting ourselves, while forcing our customers into an irrational tradeoff:

Pay more so more people can look at the same truth.

That’s not how buyers think, especially when budgets are tight and value is indirectly delivered.

Pricing didn’t just slow adoption, it changed behavior

The user-based pricing model didn’t just limit our growth. It impacted how customers perceived and used our product.

First, access became gated. A small group of “power users” controlled insights and redistributed them via decks, exports, or screenshots. The truth traveled second-hand. Funnily enough, this was exactly what our product was intended to solve!

Second, customers under-invested in scope. Once the number of licenses was discussed, teams became cautious about adding sources or expanding use cases. Not because the value wasn’t there, but because pricing framed the product as a personal productivity tool, not as an organization's infrastructure.

That was the deeper problem.

We were trying to position our product as a shared layer for customer truth for the entire organization. At the same time, our pricing was teaching customers to treat it like a personal productivity tool.

User licenses weren’t where our costs lived

If stalling growth wasn't already enough, there was a more fundamental problem below the surface.

Our costs didn’t scale with users. They scaled with work done.

• Hours of audio and video processed
• Feedback items ingested
• AI tasks executed
• Workflows run repeatedly by AI agents

You could have:

• many stakeholders who needed visibility (high value to the customer, low marginal cost to us), or
• a small team doing heavy processing (low seat count, high variable cost for us)

In one case, seat pricing felt unfair to the customer.
In the other, it was unfair to us.

Either way, the pricing model was totally misaligned with reality. And this type of misalignment is what kills businesses – slowly at first, then all at once.

What are customers actually buying?

Once we named the real issue, the right question followed naturally:

What is the unit of progress customers are actually buying?

Spoiler alert: It isn't access.

For us, progress looked like:

• customer truth processed
• insights generated
• workflows shipped
• coverage of customer data
• decisions unblocked

So instead of pricing who could look at the data, we needed to price the AI engine doing its job.

In hindsight, that sounds obvious. It wasn't at first.

The moment you say “usage-based,” procurement teams hear “unpredicable.” Predictability matters, especially in enterprise settings.

The challenge wasn’t usage vs. subscription. It was: how do you align price with "work done", without making spend feel uncontrollable?

What we tested, and what we refused to do

We didn’t flip the model overnight. We tested across segments, knowing each would punish us differently.

Three principles were non-negotiable:

1. Don’t price compute
Customers don’t want to buy tokens, model calls, or chain lengths. It's way too hard to explain, and truthfully, it's our concern, not theirs.

2. Don’t make it unpredictable
The first thing customers raised was: how do I make sure I don't end up with a big bill I didn't expect?

The last thing we wanted, is to break the trust and loyalty of our customers while making this shift. It's what got us to where we are today. We decided not to charge overages, but instead throttle input and processing, while clearly communicating where the customer is at.

3. Don’t tax distribution
If insights are meant to spread through the organization, pricing shouldn’t put toll gates on sharing it.

Eventually, the model we landed on wasn’t pure consumption. It was a system of:

• Unlimited free users
• Volume limits tied to real work done (data processed, AI tasks run, AI messages sent, etc.)
• Clear pricing tiers teams could reason about and budget for

Adoption became cheap, throughput became valuable, and that alignment changed everything.

Customer perception and behavior shifted

The first impact wasn’t a dashboard spike in revenue. It was a different kind of conversation.

Instead of customers asking:

“Can we reduce seats?”

We heard:

“What happens if we add surveys?”
“If we onboard another region, what does that unlock?”
“How much more could we do if we connect our entire sales call library?”

Those questions signaled something very important: customers weren’t optimizing for cost anymore. They were ideating use cases and imagining value before it happened.

Internally, the system spread more naturally. Product, CS, Marketing, and leadership could all engage without triggering procurement debates. The product finally behaved like infrastructure, not another personal AI copilot.

Renewals stopped being login audits

Renewals of user-based pricing models tend to turn into login audits. For infrastructural systems that create value indirectly, that’s a terrible proxy, and an easy one to attack for procurement teams tasked with saving money.

Volume-based renewals flipped the narrative.

The conversation shifted from:

“Who used this?”

to:

“In which teams and business units is this currently embedded?”

Systems that are properly used become hard to rip out. Not because of contracts, but because of dependency.

The operator lesson

For me, the biggest takeaway wasn’t about pricing mechanics. It was about leverage. Pricing doesn’t just capture value to your business. When done correctly, it shapes customer behavior and accelerates growth.

Ultimately, the changes we made didn’t just change how customers paid the bill. Instead, it changed how our product was perceived, how it was used and how we could realize value delivery.

The simplest way I can describe it:

• We stopped monetizing access and started monetizing value delivered.
• We stopped taxing distribution and started enabling it.

For AI-native products especially, pricing is not a static decision. It’s a growth loop. If you’re still pricing people while your value and costs scale with work done, you’ll eventually hit the same wall we did.

Admitting what your product really is can be uncomfortable. But once pricing reflects that truth, growth stops being something you push. It becomes something the system produces.

What’s Next?

• 🚀 Want more of this? Subscribe to get PLG & CS insights in your inbox.
• 💬 Have thoughts? Reply, share your experience or connect with me!

FAQ

Why does user-based pricing often fail for AI-native products?

User-based pricing fails when value and cost scale with work done, not with logins. In AI-native products, insights, automation, and decisions compound as usage grows. Pricing users creates adoption friction, encourages gatekeeping, and disconnects revenue from the actual value and cost drivers of the system.

What is the “wrong unit of progress” in AI pricing?

The wrong unit of progress is access. Most AI products don’t create value when someone logs in, but when data is processed, insights are generated, and workflows run. Pricing users assumes value equals usage of an interface, which breaks down for AI systems operating in the background.

What is volume-based pricing in AI SaaS?

Volume-based pricing charges based on work performed by the system, such as items processed, media minutes analyzed, AI tasks executed, or workflows run. It aligns pricing with how customers experience value and how AI infrastructure costs scale, instead of who has access.

How does user-based pricing change customer behavior?

User-based pricing doesn’t just slow adoption—it alters behavior. Customers gatekeep access, share insights second-hand, and limit scope to justify seat counts. This prevents insights from spreading and discourages system-level use, especially when the product is meant to function as shared infrastructure.

Why does pricing access discourage organizational adoption?

When access costs money, teams optimize for fewer users. That leads to centralized “power users” and downstream consumers. For systems that create value through shared truth, pricing access directly contradicts the intended behavior of broad visibility and cross-functional alignment.

Why don’t AI company costs scale with users?

AI costs typically scale with processing volume, not people. Compute, storage, inference, and orchestration costs increase with data ingested, tasks run, and workflows executed. A customer with few users can generate high costs, while many stakeholders may add near-zero marginal cost.

What should AI companies price instead of seats?

AI companies should price the unit customers perceive as progress: data processed, insights generated, workflows shipped, or decisions unblocked. The ideal pricing meter sits at the overlap between customer value and variable cost, and is predictable enough to budget against.

Is usage-based pricing the same as consumption pricing?

Not necessarily. Pure consumption pricing can feel unpredictable. Many AI companies succeed with hybrid models: predictable tiers with clear volume limits, no surprise overages, and simple upgrade paths. The goal is alignment, not metering every compute detail.

How do you avoid bill shock with volume-based pricing?

Avoid charging hidden overages. Use clear limits, visible usage, throttling instead of surprise fees, and simple upgrade paths. Customers fear unpredictability more than price itself. Trust is preserved when spend is understandable and controllable.

Why is pricing a growth lever, not just a revenue decision?

Pricing shapes behavior. It determines who adopts, how broadly value spreads, and whether customers optimize for cost or outcomes. When pricing reinforces desired usage patterns, growth becomes systemic rather than something sales and CS must constantly push.

How does pricing affect retention in AI products?

Seat-based renewals often become login audits, which is a poor proxy for value in infrastructural systems. Volume-based pricing reframes renewals around embedded usage: data processed, workflows running, and decisions supported. Systems that are operationally embedded churn less.

When should a startup consider moving away from seat-based pricing?

If customers negotiate access instead of outcomes, if adoption stalls despite strong interest, or if costs don’t scale with users, seat-based pricing is likely misaligned. These are early signals that pricing is working against, not with, the product.

Is volume-based pricing only suitable for large enterprises?

No. SMBs value simplicity, while enterprises value predictability. A well-designed volume model can serve both by keeping usage legible, tiers clear, and adoption friction low. The mistake isn’t volume-based pricing—it’s making it complex or opaque.

What’s the biggest lesson from moving from user-based to volume-based pricing?

Pricing doesn’t just capture value—it creates it. When pricing aligns with how value is produced and shared, adoption accelerates naturally, customer behavior improves, and growth becomes an output of the system rather than a constant fight.