Gavel Gavel
Experts Guides Skills
Cited from real sources 6 min read Updated July 2026

A pricing framework by Madhavan Ramanujam

Madhavan Ramanujam's AI Pricing 2x2: Attribution and Autonomy

Madhavan Ramanujam, senior partner at Simon-Kucher, maps AI pricing onto two questions: can you prove the value you deliver, and does the product work without a human in the loop? Attribution and autonomy. Where you land on those axes determines whether seat-based, hybrid, usage-based, or outcome-based pricing is the right archetype, and how much pricing power you actually hold.

The two axes

"when you have high attribution and high autonomy, that is when you have high pricing power"

Classic SaaS captured 10 to 20 percent of the value it created. Ramanujam says an autonomous, attributable AI product can take 25 to 50.

Madhavan Ramanujam on Lenny's Podcast Pricing your AI product Watch at 39:23

The framework

Two axes, four archetypes

The question every AI founder asks is which model to use: seats, usage, or outcomes. Ramanujam's answer is that the model is not a preference, it is a readout of two properties your product either has or does not.

Attribution is whether you can prove the value you created, ideally against a KPI the business already tracks. Autonomy is whether the work happens without a human in the loop. Plot those and the archetype falls out, along with how much of the value you can defensibly claim.

Low attribution · Low autonomy

Seat-based or subscription

You cannot attribute much and there is a human in the loop. Charge per seat, then work on attribution to move right.

High attribution · Low autonomy

Hybrid

Copilot products like Cursor. Keep a seat fee for the assist, layer credits or tokens on top so heavier use pays more.

Low attribution · High autonomy

Usage-based

Backend and infrastructure that runs unattended but does not move a tracked KPI. Seats make no sense with no human; usage proxies the value.

High attribution · High autonomy

Outcome-based, the golden quadrant

Intercom's Fin charges per AI resolution and nothing when a human steps in. ChargeFlow takes a cut of chargebacks it recovers.

Ramanujam's numbers on where the market actually sits: hybrid is the most common model today, roughly 5 percent of companies run true outcome-based pricing, and he expects that to reach about 25 percent within three years. The companies already there capture 25 to 50 percent of the value they create, against the 10 to 20 percent that was considered good in classic SaaS.

How to apply it

How do you pick your AI pricing model?

Six moves, from how Ramanujam tells founders to use the 2x2.

  1. 1

    Place yourself honestly, today.

    Find your current quadrant before picking a model. Rushing to outcome-based without provable attribution is the failure mode he names first.

  2. 2

    Price from day one, not after traction.

    AI cost dynamics and attribution are decided early. Retrofitting monetization onto a product trained on a cheap price is far harder than starting right.

  3. 3

    Do not anchor low to buy market share.

    Training buyers to expect $20 a month sets a ceiling you then have to escape with a separate, higher-priced product. Growth without wallet share is fragile.

  4. 4

    Build attribution deliberately to move right.

    If you are in the seat-based quadrant, the work is instrumenting proof against a KPI the customer already tracks. That is what buys pricing power.

  5. 5

    Add autonomy to move up.

    More agentic, less human-in-the-loop moves you toward the golden quadrant. Autonomy is what justifies charging for work delivered rather than access granted.

  6. 6

    Charge for the outcome once you can prove it.

    Price against resolutions, recoveries, or hard savings. Charge nothing when a human has to intervene, which is what makes the claim credible.

you started training your customers to expect $20 a month and you anchored yourself on a low price point
Ramanujam on the underpricing trap in AI tooling Watch at 37:59

Boundary conditions

When it works, when it fails

Works best when

  • The AI touches a metric the customer already tracks, so attribution is arguable
  • You can point to hard savings or recovered revenue, not just time saved
  • The work completes without a human, which is what justifies outcome pricing
  • You are setting pricing early, before buyers are anchored on a low number

Fails when

  • You jump to outcome-based pricing without being able to prove attribution
  • The product still needs a human in the loop, so the outcome is not really yours
  • You chased market share with a cheap price and now cannot raise it
  • Fast revenue growth hides churn, because the customers were never retained

The failure Ramanujam keeps returning to is growth that is not enduring. Revenue can climb quickly on a price that was set to win logos, and still leave a business that cannot raise, cannot retain, and never captured the value it created.

if you just threw out a $20 product hoping to just, you know, accelerate your market share, you're in trouble
Ramanujam on market share without wallet share Watch at 38:37

Where operators disagree

Ramanujam's instinct is to price against provable value and treat a low anchor as a trap. Alex Hormozi pushes the same direction harder on the human side with the gasp test, arguing you have not gone high enough until saying the number is uncomfortable. The counter-case in AI tooling is deliberate underpricing to win the category first, then expanding wallet share later. Ramanujam does not reject that outright; his condition is that it only survives if there is a real land-and-expand plan, not just a hope that price can be fixed later.

This is the AI-era companion to Ramanujam's older rule that 20 percent of features drive 80 percent of willingness to pay. That one tells you what to charge for; this one tells you what shape the charge should take once the product does the work itself.

The sources

Where Ramanujam discusses this

36:50
Timestamp

Lenny's Podcast · Pricing your AI product

The AI Pricing 2x2 Framework

The pricing power of an AI product is determined by its level of autonomy and its ability to prove value attribution. While hybrid pricing is currently the most popular model, the ultimate goal is to move toward outcome-based pricing.
27:36
Timestamp

Lenny's Podcast · Pricing your AI product

Unique Challenges of AI Monetization

AI products require monetization strategies from day one due to distinct cost dynamics and the ability to solve attribution problems. Founders must move away from pay-for-access models.
48:03
Timestamp

Lenny's Podcast · Pricing your AI product

Executing the Path to Outcome-Based Pricing

Companies should assess their current position on the 2x2 framework and plan an evolutionary path toward outcome-based models by enhancing attribution mechanisms and building more agentic, autonomous features.
11:46
Timestamp

Lenny's Podcast · The art and science of pricing

Willingness to Pay Conversations

The core of pricing strategy is testing willingness to pay before building. Madhavan shares methods like relative value indexing and the acceptable/expensive/prohibitive price framework.

Useful? Pass it to a founder about to launch at $20 a month.

Want the full playbook?

Get 98 product management frameworks.

31 frameworks 12 rules 47 heuristics & principles 49 operators

From Chandra Janakiraman, Stewart Butterfield, Ami Vora, and 46 more. Drop one .md into Claude, Cursor, or ChatGPT. Your AI cites practitioners, not guesses.

See the pack

Instant .md download · One-time purchase · No subscription

Related frameworks

Pricing

Willingness-to-Pay 20/80 Axiom

Madhavan Ramanujam

20% of features drive 80% of willingness to pay. Most founders give that 20% away free.

Pricing

Van Westendorp Pricing

Rahul Vohra

How Superhuman priced at $30/mo using the Van Westendorp price sensitivity meter: four questions, real buyers.

Pricing

Gasp Pricing

Alex Hormozi

If the buyer doesn't gasp at the price, you didn't go high enough. Anchor, then walk down.

Positioning

Beachhead Market Strategy

Geoffrey Moore

Big enough to matter, small enough to lead, a good fit with your crown jewels. How to pick the one segment you commit to.