software pricing has changed shape seven times in 60 years, each shift happened because the previous model stopped capturing the value of the new technology. 

we're mid-way through shift number seven right now, and most vendors are picking the wrong metric for the wrong reasons.

the short version tl;dr:
consumption-based pricing is the current market default for AI. it's also the second-best model, and while outcome-based pricing is the right answer, but it requires solving attribution.

and honestly, the attribution is the hardest unsolved problem in AI monetization, and we're focusing so much on that!

all insights are mine, no AI slop, even though I am talking about LLMs and stuff - even this description is manually edited, crafted, and polished by myself - o tempora o mores, where we are as a world we actually need to say it...

this episode walks through the whole history of software metrics, but with a twist on which metrics to actually pick and when:

• the seven pricing shifts, from mainframe hourly rates to outcome-based agents
• why per-license pricing worked in the PC era and broke when cloud hit
• the birth of SaaS tiers and how "customer success" became a job title
• seat-based pricing as the accidental default that lasted 20 years
• usage-based pricing and value metric picking (messages, mentions, keywords)
• why AI vendors reached for consumption pricing first - and of course why customers accepted it
• token-based pricing and the margin exposure problem when model costs drop 80% a year
• output-based vs outcome-based: they're not the same thing, and you should kniow it!
• resolution pricing at Intercom, recovery pricing at Chargeflow - few examples I believe should be here
• attribution as the wall every outcome-based startup eventually hits
• pick second best hypothesis: why software always picks the workable model first and the right one later
• aconcrete framework for choosing your pricing metric in 2026

solo-engineered by Maciej Wilczynski, Ph.D., Managing Partner at Valueships, always below 20 minutes.

---------

timestamps

00:00 intro — the pricing question every AI founder is asking 
01:30 mainframe hourly rates: the original usage model 
03:10 PC era and per-license pricing 
04:30 cloud computing and the birth of SaaS tiers
05:30 how subscriptions created "customer success" as a function 
06:30 seat-based pricing and the value metric era 
08:00 why AI reached for consumption pricing first 
09:10 the token cost problem: 80% price drops don't mean 80% price cuts 
10:30 output-based pricing and the mid-tier compromise 
11:30 outcome-based examples: Intercom, Chargeflow
13:00 the attribution wall - how to overcome it in a right way
15:00 second best hypothesis: why software adopts workable before right 
16:30 consumption as the current default 
17:30 where outcome-based pricing actually works today 
18:50 how to pick your metric in 2026
---------

key takeaways:

consumption pricing is the market's current answer, but not because it's the best model. simply it's the one that is actually managable, vendors can implement it, customers can accept it, procurement doesn't fully hate it, so it's a trade-off no one really wants, but that's the ad reality.

token-based pricing has a margin problem, which will be a problem in the future. foundation model costs are dropping 60-80% per year. to put in perspective: if you priced your product on 2024 token economics and customers now expect that pricing to hold, you're either eating margin compression or renegotiating downstream - both are bad.

outcome-based pricing is the future, but only where attribution is clean. Chargeflow can price on recovered chargebacks because every recovered dollar is measurable and directly attributable, while Intercom charges for resolution - only when you have clear, clear attribution you can actually get it right. outcome-based pricing doesn't work in broad use-cases.

second best hypothesis: software always picks the workable model first, not the right one. SaaS didn't launch with per-outcome pricing, but with per-seat because that was the easy operational model. same story now: consumption before outcome, because consumption is what founders can ship and customers can budget for.
---------

for your own product in 2026, the framework is:

1. dollarize the value - if you can't put a specific dollar figure on what your AI delivers per user, you can't price on outcomes yet
2. solve attribution - can you draw a straight line from your product's action to the customer's business result?
3. assign ownership - who at the customer's org owns the outcome and can approve the pricing
4. show confidence intervals - customers accept outcome pricing when you can predict impact with a range, not a single number
5. protect your margin - model costs will keep dropping; your pricing structure needs to survive that

a) If you can hit all five, price on outcomes and charge premium
b) If you can hit three, price on outputs and charge fair. 
c) If you can hit fewer, price on consumption and don't apologize for it, that's the workable model until the market gets smarter.

referenced in this episode

• Intercom,a resolution-based pricing for support tickets
• Chargeflow, a recovery-based pricing on chargebacks
• Related Valueships reading: The Real Economic Value of AI · AI Pricing services

---------

frameworks referenced

• Second Best Hypothesis — my own framing, expanded in this episode
• The AVI (Artificial Value Index) — introduced in Episode 1, deep-dive coming later
• Valueships AI SaaS Pricing Canvas — Krzysiek Kobylecki's 8-element framework, full breakdown in a future episode

New episodes every Wednesday. Listen on Apple Podcasts · Spotify · or right here.

If this resonated, hit subscribe. If it didn't, tell me why — I read every reply.

solo-engineered by Maciej Wilczynski, Ph.D., Managing Partner at Valueships, always below 20 minutes.

went through what Goldman, McKinsey, MIT, Stanford, and Acemoglu have to say about AI productivity gains and also considered what it means for pricing. short version: AI bolted onto existing processes caps at 15-30% productivity gain. 

that's not enough for a pricing premium. you need to rebuild your work around AI as Henry Ford did. This is the way to unlock real value. 

also mapped out 5 situations where AI actually justifies a price premium.

keeping it short under 20 minutes, I value your time. 

whole report I quote is here, without any e-mails or login: https://www.valueships.com/artifacts/the-real-economic-value-of-ai

timeframe:

• 00:00 intro
• 00:28 67 reports on AI productivity ceiling
• 02:13 the Henry Ford parallel - you need to make your own workflow
• 03:13 theory of constraints vs. AI
• 04:26 the Artificial Value Index - how it works
• 05:30 MIT study on how most pilots fail
• 06:01 the pricing model mismatch
• 07:47 why customers stick to old purchasing models
• 10:08 five situations where AI justifies a premium - good checklist to follow
• 13:13 cost vs revenue framing
• 14:13 closing - AI is the anchor, not the keynote

Maciej Wilczynski, Ph.D. 
Managing Partner, Valueships