Reforge’s Brian Balfour on Why AI Breaks Standard SaaS Monetization (and Why That’s Okay)

TL;DR

In this Second Acts episode, Reforge CEO Brian Balfour joins Chargebee’s Krish Subramanian to explain why old SaaS pricing assumptions are breaking, the three tests every monetization model must pass, and why pricing has become an ongoing, hypothesis-led experiment.

As Krish summarized on LinkedIn: “AI monetization completely breaks traditional SaaS models such as freemium… Frequent, hypothesis-led pricing iterations are thus really important.”

Balfour echoed the same point in the conversation:

“You used to do maybe a pricing change once per year. Now it’s three, four, or even five changes in a relatively short period just to keep up with shifting costs and adoption.”

Classic SaaS models worked because serving one more user was essentially free, which made freemium, per-seat pricing, and viral loops possible. With compute-heavy AI workloads, every query, workflow, and experiment now carries a real cost — and value is shifting from simple seat access toward measurable outcomes. 

Pricing, Balfour argues, has always been a multi-part puzzle, but today those puzzle pieces are moving faster and in more complex ways than ever before.

Why old SaaS pricing models are under pressure

In traditional SaaS, free users were nearly costless. In AI-driven SaaS, they’re not.

Balfour explained that assumptions about negligible marginal cost no longer apply. Models like freemium and per-seat break down when every interaction burns compute.

Imagine offering a free plan where each query costs you $0.01. Ten thousand free users running 100 queries per month come out to $10,000 in infra spend before you’ve earned a cent. That’s why “old” playbooks can quickly sink new companies.

The pricing spectrum: seats → usage → outcomes

Balfour described pricing along a spectrum:

• Seat-based: still common in enterprise SaaS; simple to buy, but brittle when actual engagement varies.
  
• Usage-based: increasingly popular in infrastructure; aligns with value, but requires clear guardrails and customer trust.
  
• Outcome-based: vendors assume more risk to capture more value. This is still rare, but some industries (CX, fintech) are experimenting.

Balfour observed that most companies are already moving from seat-based to usage-based pricing. Outcome-based models remain rare, yet their appeal is clear: They let vendors capture more of the value they create, even if it means taking on more risk. 

Each rightward step along the spectrum increases the vendor’s risk but also expands the share of value they can capture.

The three tests every monetization model must pass

Pricing, he argues, is a puzzle with four pieces:

The puzzle only works if it passes three tests.

1. Customer value test
   
   “When the customer looks at your price, the question they’re really asking—often not rationally, but emotionally—is: does the perceived value exceed the perceived price by a multiple?”
   
   
   Without that, adoption stalls.
   
2. Growth model test
   Pricing must enable acquisition loops, not choke them.
   
   He cites Miro as an example: Their free plan initially capped the number of users. Teams began deleting colleagues to stay under the cap instead of upgrading. By shifting the limiter to boards instead of users, Miro unlocked virality while improving monetization.
   
3. Cost-to-serve test
   
   The monetization model has to sufficiently support your cost to acquire and the  cost to serve customers.
   
   In AI-driven products, the cost-to-serve test has become especially important. As Balfour highlighted, drawing on Ethan Ding’s (co-founder and CEO, TextQL)  AI Subscriptions Get Short Squeezed, older AI models get cheaper as they mature. 

However, the best-performing new models stay expensive, and because more complex use cases consume more tokens, the actual cost of serving customers is actually climbing.

Monetization lags demand in every platform shift

Balfour compares today’s AI moment to the early years of Google search. 

At the time of its Series A, Google’s monetization “plan” was three-pronged: sell enterprise search, experiment with banner ads, and license results to portals and directories. None of those became the real business. 

The breakthrough — self-serve, pay-per-click ads — came later with AdWords.

AI, he argues, is in a similar phase today. Demand and value are undeniable, but the dominant monetization approaches are yet to emerge. Companies are still testing different forms of pure-play and hybrid models.

“I think we’re very much in the figuring-it-out part of the cycle.”

Don’t discount today’s outliers

That’s why Balfour cautions against declaring winners and losers too quickly. Monetization models are still emerging, and companies experimenting at the edge often shape the standards that follow. Cursor, for example, has drawn attention as one of the fastest-growing AI startups.

Ariela Bitran, Chargebee’s Head of Pricing, argues that companies like Cursor should resist cost-plus approaches and instead orient around value pricing — tying price to the outcomes created, not the compute consumed. 

It’s a perspective that underscores Balfour’s point: the long-run winners will be those who innovate not just in product, but in how they capture value.

Inside Reforge’s AI-native pricing experiments

Reforge is building an AI-native suite for product and growth teams. 

The suite currently includes Insights (priced on data volume, with unlimited seats), a Research product, and a Prototyping product. That context shapes how they think about pricing each element individually, while keeping the overall suite coherent.

Current experiments include:

• Insights: unlimited seats, priced on data volume/usage
  
• Research: pricing still under development, exploring models that balance simplicity with adoption
  
• Prototyping: testing hybrid seat + usage models
  

Balfour is clear that outcome-based pricing isn’t a fit yet:

“None of these products are very well suited for outcome-based pricing yet, so we’re not going there.”

What’s consistent is the philosophy: pricing is treated as a living system of hypotheses.

Beyond product-level pricing, Reforge is also working to ensure the suite’s models are cohesive: “We don’t want to make the pricing so different from product to product that it’s impossible for each person to comprehend.

Why SaaS companies need pricing infrastructure now

The biggest shift? Pricing now changes multiple times a year.

“You used to do maybe a pricing change once per year. Now it’s three, four, or even five changes in a relatively short period just to keep up with shifting costs and adoption.”

That makes infrastructure indispensable. Companies need systems that can:

• Meter and track usage cleanly
  
• Manage entitlements and packaging
  
• Run rapid experiments on value metrics
  
• Update billing and communicate changes without breaking customer trust
  

Without this layer, pricing agility becomes a bottleneck. With it, pricing can become a competitive moat. And, as Balfour notes, that’s precisely the kind of infrastructure Chargebee is obsessed with enabling — giving SaaS companies the flexibility to experiment, adapt, and scale their monetization models as quickly as the market shifts.

What’s next: retention, teams, and capital

Balfour’s core message: treat monetization as an ongoing system of hypotheses, tests, and adaptations.

In our next post, we’ll cover the rest of Krish and Brian Balfour’s Second Acts conversation:

• Why retention is fundamentally about frequency, not features.
• How small teams ship absurdly fast.
• The pain and payoff of incubating a startup inside a larger company.
• How founders should think about capital signals in hype cycles.

Listen to the full episode on your favorite podcast platform below:

Even more insights

Discover lessons from the field on using pricing to drive adoption, build sticky habits, and create an unfair advantage in AI markets.

Get the Pricing for AI Distribution Playbook →