Growing Ventures
AI Made Software Easier to Build But Harder to Price
Not too long ago, I worked with a client on a new AI-native software product that helped their users create personal branding content for social media. Their main business, however, is a hands-off service where they produce personal branding content for their customers, yet some customers wanted to be more hands-on themselves. This product was for them.
What was interesting though is not the product that we built, but the pricing model and the value proposition that we applied.
They charged a four-figure fee per month for access to both their AI-native software and personalized consulting from their team.
What they really sell – on top of the AI-native software product – was expertise, which I think is a genius way to avoid the trap many founders and operators currently fall into: reselling tokens with a markup.
How SaaS used to work
For years, the Software-as-a-Service (SaaS) playbook was simple: Build software once, then charge per seat and cash in gross margins of sometimes up to 85%. The concept behind these unit economics is called "(almost) zero marginal costs".
You build out your software business once with higher costs upfront, then have customers pay for it. It eventually does not make much of a difference if you serve 1,000 users or 10,000 users through your software product. Costs would not scale linearly with revenue.
But that playbook is about to change.
AI-native businesses are built on different unit economics
Let's take the AI app builder Lovable as an example. Their business revolves around using other companies' foundation models – think OpenAI, Anthropic or Google – wrapping them in a well-designed user experience and charging a markup on the tokens.
They are in the reselling tokens business.
Every prompt their customers generate costs them money: real, variable, per-customer money. These are not classic SaaS economics anymore, but closer to running a hosting service with a brand on top.
But what is the alternative for AI-native businesses?
Your cost structure and value creation are your pricing moat
Seat-based pricing is also breaking. Take companies like Intercom, Sierra and Zendesk for example who have moved to per-resolution pricing, because the seat model penalizes them every time their AI gets better.
Salesforce now runs three pricing models in parallel, because it cannot fully replace seat revenue without cannibalizing itself.
I think founders who are still pricing their AI-native businesses per seat are working from an outdated map. The math of SaaS is being rewritten and the pricing models needs to follow cost structure and value creation, not the other way around.
What I'd suggest to founders and operators
The business fundamentals are changing and you need to adapt your business accordingly.
- Tailor your pricing strategy: Decide early whether your gross margin looks more like SaaS or more like infrastructure. If you are simply wrapping an AI model, consider adding value add on top: services, community, deep workflow integration, moats around proprietary data and more.
- Watch where pricing is moving in your category: Outcome pricing is real, but only works when your unit economics support it.
- Ask yourself where your real value sits: If the next foundation model upgrade can replicate what you do, you are not building a software business – you are building a feature.
I'm still figuring out how this refines our acquisition strategy and how our portfolio businesses will need to adapt in an AI-native future, but one thing is certain: this future is coming sooner than we might expect.
How are you adapting to it?