Hello, pricing folks!

Time machine. We're going back to the early 2000s (or should I say Y2K, six seven, or being completely delulu?) when the first cellphones were a thing, Neo was saving us from the Matrix with his foldable Nokia.

While we were all pumped by the devices in a far more technocratic and enthusiastic way (well, it was far before some tech moguls decided to increase our dopamine and anxiety hits to boost their metrics, another story though), there was an important pricing phenomenon we would really like to forget, and yet it comes back to us. 

Remember O2, Vodafone, AT&T, Orange (or any other provider) overcharges? So, effectively, you had a plan of minutes and text messages, and if you went beyond it, you were charged. A LOT.

We hated it so much at that time. The key reason they were doing it, beyond pure greed, was that, at the time, the TelCo market lacked a clear standard for connecting with one another.

To put it simply, they were all using different currencies and were charging each other for making a transaction call. The customer paid the spread. Otherwise, their margins got squeezed, they would sell below the bottom line, and it wouldn't work. For TelCos, it was a way to control the costs and ensure they have a decent pocket margin.

The thing is, what we truly hated was not the consumption itself. That felt true and honest. Pay-as-you-go, I use, I pay, and all these. The problem was when you didn't really have control over it. This is what really drove us nuts. 

However, once the market is established, some disruptors (like P4) already know how many minutes and text messages people use, so they package it all into "unlimited" offers, which cover 90-95% of use cases. Pricing pressure increased; overall margins were squeezed; operators moved towards more value-added services (VAS); and others did so because it was hard to make money on pure consumption.

This is AI now, and we need to wait until someone packages it rightly, so we get rid of overages and potential problems with someone burning half of the AI budget in a day.

As for now, as customers, we still have PTSD that creates a huge blockade to usage-based adoption. It's not about consumption, but uncontrolled spend. 

It's a trust issue, that's why it will be hard for us to go towards outcome-based pricing, which is a truly Holy Grail of pricing, and I have some data to prove it:

As you can see in the image above, coming straight from our "Real Value of AI" report, enterprise companies prefer consumption. Also, they are pretty much the same, almost in par, with having a flat platform or old school license-based fees.

What they are still not ready for is: "outcome-based pricing". Whenever a Valueships client comes to me, I ask: "Do you really believe your customers want it?"

This is literally an overages charges problem dressed as something else entirely. 

If you can't attribute the value, you don't know who is responsible for the output/outcome, how can I even bill it properly? 

It may happen, but you need to have two economic value rules checked out:

1. You can prove real, quantified value, and your client is aligned on that. Real numbers in dollars, not some mambo jumbo BS.

2. You have a clear 100% success attribution; in other words, you take care of your client problem entirely.

From my perspective, on top of the questions above, you can do this easy 10x pricing test for outcome-based pricing. You can charge for outcome based if you meet at least one of these 5 conditions:

1. Scarce specialist labour replacement - senior radiologist, M&A partner, ML scientist - when AI quality reaches that level on the specific task. Look at the new Fable model, it's quite capable, so we're close. Don't even think of outcome-based if you're replacing interns, obviously. 
   
   
2. Revenue-generating outcome with clean attribution -  conversion lift, fraud recovery; where procurement accepts the attribution. You see? This is essential, someone needs to accept it, and sign out the whole thing. 
   
   
3. Bottleneck / capacity unlock - drug discovery cycle, M&A diligence, where time-to-market value dwarfs the price. Check recent OpenAI's developments in pharmaceutical cancer research; crazy progress justifying the high costs.
   
   
4. Regulated workflow with liability - AML/KYC, SOX, medical documentation - insurer-style premium where AI assumes risk. Check 11labs and their insurance policy - they're literally tapping on that, and trying to bill on outcomes.
   
   
5. Mission-critical reliability with human in the loop - all the other narrow domains where the alternative is unacceptable. So Palantir-like models where AI does the heavy lifting, yet the final decision comes to human. Still attribution and value is clear though.

You can read the whole report on AI value here:

Take care and enjoy! If you reply, I will personally receive this e-mail.

I am absolutely amazed how many people write back to me - that's why I'm doing it. And of course for the money. 

Maciej Wilczynski, Ph.D.

Managing Partner