Why your old SaaS playbook is killing your AI startup

This article summarizes a an episode of Lenny’s Podcast featuring Madhavan Ramanujam, managing partner at 49 Palms Ventures.

49 Palms Ventures General Partner Madhavan Ramanujam / Image credit: First Round Capital

The easiest 20% of an AI product to build often drives 80% of a customer’s willingness to pay. In a discussion with Lenny Rachitsky, 49 Palms Ventures managing partner Madhavan Ramanujam argues this paradox forces AI companies to master monetization from day one, a stark departure from the traditional SaaS playbook.

This mandate demands a strategic re-evaluation of pricing, which Ramanujam organizes into a 2×2 framework based on a product’s autonomy and attribution. This analysis reveals how the framework provides a clear pathway for AI founders, from choosing an initial model to executing a proof-of-concept designed for value capture.

The four pricing archetypes

The 2×2 framework plots pricing models based on a product’s autonomy (how much it operates without human intervention) and its attribution (how clearly its value can be measured). Each quadrant corresponds to an optimal pricing archetype that aligns with the value it delivers to the customer.

Low Autonomy / Low Attribution: The Seat-Based Model
Ramanujam explains, “The quadrant where your attribution is low and your autonomy is low, the best pricing archetype that actually fits is a seat-based or a subscription model because there’s not much to do about it”.

An example is a product like Slack, where it’s understood that productivity increases, but that value is difficult to precisely measure and attribute back to the software.

High Autonomy / Low Attribution: The Usage-Based Model
He details, “These tend to be mostly like backend or infrastructure products that are core critical to run businesses…in that situation you need to be on a pay for what you consume, and a usage based model [makes sense]”.

Examples of this model include AWS, which charges based on compute usage, and OpenAI, which charges per token.

Low Autonomy / High Attribution: The Hybrid Model
Ramanujam notes, “You still have a seat-based model for the co-pilot kind of use case, but you also layer in a consumption model which actually says there are a certain number of AI credits or tokens that can layer in the usage aspects”.

Companies like Cursor and Canva fit this model. For instance, Cursor clearly improves developer productivity (high attribution) but still operates in a co-pilot mode with a human in the loop (low autonomy).

High Autonomy / High Attribution: The Outcome-Based Model
He identifies, “The quadrant that you really want to be in is the golden quadrant which is the top right one. That’s the outcome-based pricing model where you have great autonomy and great attribution”.

Examples include Intercom’s Fin, Sierra, and ChargeFlow. Fin exemplifies this by charging per AI resolution, meaning it only charges for a successful outcome that occurs without any human intervention.

The AI pricing mandate is fundamentally different

AI companies cannot afford to treat monetization as an afterthought. Unlike previous SaaS models, the inherent cost of goods sold and the scale of value delivered demand immediate financial validation. Founders must abandon old playbooks or risk training customers to expect immense value for minimal cost.

The golden quadrant is outcome-based pricing

Weaponizing the Proof of Concept