wrote in 2019 A post on how companies should price their AI-enabled software. I focused on his SaaS company developing its own AI and focusing on pricing considerations when working to improve the model.
Since then, there has been a rapid rise of third-party base model providers such as OpenAI, MosaicML. These “His AI as a Service” vendors have enabled any His SaaS player to integrate powerful AI into their applications. This has created a frenzy of AI fairy dust being sprinkled across the SaaS ecosystem. We have seen this in countless newly formed start-ups and more established public companies.
The pervasiveness of this technology raises many questions, such as how to deploy it safely and who will win – a rising startup or an incumbent with an existing distribution? One key area that hasn’t been discussed much yet is how to set the price.
Below is a working framework on how to think about pricing AI for SaaS applications. This universe is evolving rapidly, so I will update on this idea in future posts.
How much differentiated value do AI capabilities create?
By definition, these base models are accessible to all SaaS providers.So how should we think about pricing for products that are actually integrated into the product? Let’s start with the first principle: How much? differentiated What value does this AI capability create?
Integrating AI capabilities into the broader platform’s flow saves users time leaving the flow and navigating to underlying models (such as ChatGPT). Keeping users contextual is a powerful unlock.
But be honest with yourself about how much value AI is actually creating. Many AI capabilities in SaaS are now inundated with early interest from curious users, but have not seen meaningful continued adoption. Start by understanding retention and value creation.
SaaS companies need to solve for simplification of pricing and adoption of AI capabilities. Now is the time to learn and iterate.
Next, ask yourself how differentiated your AI product is. If most of the value your AI feature creates comes from going directly to ChatGPT, don’t try to get a big margin on that feature. Reselling is not a sustainable value creation strategy (nor is it a differentiating strategy, but more on that in another post).
Even if you can’t charge a lot for AI capabilities today, you can still create meaningful value by making your current products more valuable, and in some cases more attractive. It can also be used to drive upsells to higher tiers, all of which lead to better net dollar retention.
Over time, we’ll be able to build more differentiated value by leveraging the initial capabilities that are currently just thin wrappers around third-party models (we’ll show you how later). Once you reach that point, you can consider pricing approaches that extract more value.
