The Attribution-Autonomy Pricing Matrix
Pick your AI pricing model by plotting value attribution against product autonomy, then climb toward outcome-based.
- Difficulty
- Moderate
- Time to result
- ~months to results
- Steps
- 4
- Confidence
- 95%
A 2x2 that maps AI products on two axes — how clearly value is attributable to your product, and how autonomous it is (human-in-the-loop vs. fully autonomous) — and prescribes the right pricing model for each quadrant. High attribution plus high autonomy equals maximum pricing power. The framework is both a diagnostic (where am I today) and a roadmap (how do I build toward outcome-based pricing).
Origin
Developed by Madhavan Ramanujam and Simon-Kucher, drawn from work with 400+ companies and 50+ unicorns, and published in his book Scaling Innovation (sequel to Monetizing Innovation).
Core principles
- 01Pricing power comes from the intersection of high attribution and high autonomy.
- 02How you charge matters more than how much you charge, because the underlying business model has shifted from paying for access to paying for work delivered.
- 03You cannot leap to outcome-based pricing if you cannot prove attribution — pick the archetype that fits where you are today, then build toward the golden quadrant.
- 04In classic SaaS you could charge 10-20% of value; in autonomous AI you can charge 25-50% because there is no human in the loop.
How to run it
- 1
Score your product on attribution
Determine whether the value your product creates can be directly measured and attributed to it against KPIs the customer already tracks (throughput up 10%, scrap down 5%, churn reduced). Low attribution means you can only claim vague productivity gains you cannot monitor.
Pro tip Build dashboards and productized value-attribution mechanisms tied to your customer's own KPIs so attribution stops being a claim and becomes evidence.
- 2
Score your product on autonomy
Determine whether a human is still in the loop (co-pilot) or the AI completes the work independently. Backend/infrastructure tools can be autonomous but hard to attribute; co-pilots are attributable but not autonomous.
- 3
Place yourself in a quadrant and adopt its native model
Low attribution + low autonomy = seat/subscription. High attribution + low autonomy (e.g. Cursor) = hybrid (seats plus consumption/AI credits). High autonomy + low attribution (backend/infra) = pure usage-based. High attribution + high autonomy = outcome-based.
Pro tip As of recording, the hybrid model is most common because SaaS companies layered AI credits onto seat pricing; only ~5% are truly outcome-based.
Watch out Do not rush into outcome-based pricing before you can prove attribution — you will fail.
- 4
Build a vision to reach the outcome-based quadrant
Deliberately invest in more attribution (KPI-linked features, value audits) and more autonomy (agentic workforces that remove humans from the loop) so you migrate toward the top-right, where you charge for AI-delivered outcomes with no human involvement.
Pro tip Ramanujam predicts the ~5% of companies in true outcome-based pricing will grow to ~25% within three years — building toward it early is a first-mover advantage.
In the wild
Intercom historically priced customer-support software per agent (seat-based). They built Fin, which resolves support tickets autonomously, and charge only when the AI resolves a ticket end-to-end without a human; if a human intervenes, they do not charge. Fin costs 99 cents per resolved ticket.
→ Moved from seat-based into the outcome-based quadrant, pairing 'beautifully simple' pricing with attributable, autonomous value.
Cursor clearly improves coding productivity and reduces time-to-code (high attribution) but remains a co-pilot with a human in the loop. The prescribed model is hybrid: a seat-based fee for the co-pilot use case layered with consumption-based AI credits/tokens.
→ Illustrates the bottom-right quadrant where a hybrid model captures both baseline access and usage.
Common mistakes
Reaching for outcome-based pricing without provable attribution
If you charge for outcomes but cannot demonstrate the value is caused by and measurable through your product, customers will not accept it and you will fail.
Defaulting to the old SaaS seat-based playbook on an agentic product
Agentic AI taps labor budgets, which are roughly 10x software budgets; using seat-based pricing under-monetizes from day one and trains customers to expect more for less.
Is it for you?
Best for
AI founders (especially seed/pre-seed) choosing a pricing model who must decide how to charge before the model calcifies.
Not ideal for
Mature non-AI businesses whose value is genuinely diffuse and unattributable, where a simple subscription is honest and adequate.
From the transcript
“So there are two axes here. One is attribution and the other one is autonomy. And when you have high attribution and high autonomy, that…”
“The quadrant that you really want to be in is the golden uh you know quadrant which is the top right one. That's the outcome…”
“about 5% of companies are probably in a true outcomebased pricing model uh you know as of as of today but those companies some of…”
“So if you want to win in AI, figure out a way to get to that quadrant because that's the magic quadrant.”
From the episode
Pricing your AI product: Lessons from 400+ companies and 50 unicorns
Madhavan Ramanujam