How to price early stage AI and SaaS products

Welcome to Seed to Sequoia, the newsletter for founders on GTM, strategy and building companies. We’re excited to host an event at SF Tech Week–How to Get your Company Acquired. We’ve partnered with IPG and Skyp to host it in the heart of SoMa. Space is very limited, so if you’re interested please sign up ASAP! Hope to see many of you there!

And now, onto one of my favorite topics: early stage pricing.

It’s easier than ever to build and launch a software product today. The tricky parts remain figuring out how to get people to use it, and what to charge for it. Pricing is particularly fraught now that it seems like two smart people armed with Cursor can build almost anything.

How AI has changed SaaS pricing

AI-enabled products can deliver a lot more value than SaaS without AI. Therefore it can capture more value. How much more? If you look at the relationship between SaaS and one-off license-based software, SaaS captures about 1.5-2x more value. So it makes sense that AI enabled products might capture 1.5-2x more value than their SaaS equivalents.

Looked at another way, if in the old days, using software licensed from Oracle and set up yourself creates $100 million in value, Oracle might capture maybe $5 million of that value. Because the customer has to do a lot to get that value–buy servers, set up a datacenter or rent colo space, and probably integrate that with their other systems. They might even pay $5 million for a few years before ever seeing any value from the software.

SaaS changed that. Salesforce, for example, made it so the customer didn’t have to do most of that work. Just configure and get value (in theory). As a result the equivalent SaaS might get $10 million of that $100 million in value, 2x more than the legacy Oracle. Humans are still doing the work–sending the emails, taking the sales calls, updating the CRM, etc. But it’s a lot faster to value and the vendor does a lot more while the buyer does a lot less.

AI takes that one step further, because it can accomplish some of those tasks. It can take notes and update the CRM. It can write the emails (see also, Skyp). It can even pick up the phone. The customer does less, the AI does more. As a result, if its use generates $100 million in value for the customer an AI tool might capture $20 million of that value.

But how much value? How do you structure it? Those problems haven’t changed. If anything, they’re more difficult as the landscape is changing so quickly and it can be hard to explain why a customer should pay $20 million to replace a tool that they had been paying $10 million to use.

Here are some tactics and structure for figuring out pricing. It turns out this is a big topic, so unless I get a lot of emails that ask to not hear more about pricing there will be multiple posts in this series. This first post covers the principles and tactics to setting prices for all markets.

Take PMF for granted

Well–don’t, actually, but for this exercise let’s assume you have built something people want. This is a major assumption. Do the work to make sure someone actually wants what you’re building. You’re ready for this when a prospect who sees your solution (in a demo, trial or pilot) tells you that it will solve their problem, and asks what it costs or asks how or when they can start using it.

Pricing Principles

Pricing can feel overwhelming. You can literally charge anything. Palantir has billion-dollar contracts. The App Store supports prices down to $0.99. How do you decide where to be on an infinite continuum?

In setting prices, there are three key principles that can make it easier to assess both strategies and actual prices.

1. Value Alignment — Price is anchored by delivered value.

2. Fairness — Customers need to feel the price is justified.

3. Simplicity — Use the simplest model that maximizes your goals.

What matters most is the perception of your customer–at all levels, from B2C to large enterprises. If that $0.99 app seems like a good value, great. People will buy it. Same with the billion dollar team of forward-deployed AI engineers from Palantir. If the customer perceives they are getting a value from what they pay you, they will buy. It may be more than they want to pay, but it is (by definition) something they can pay.

They must also consider the pricing “fair” – which is less rational. Fairness exists both within the scope of your own pricing, and across the market and competition. How many banks advertise “Free Checking”? It seems unfair to customers charge people to maintain a checking account. The fairness principle does not have to be rational; the only thing that truly matters is the perception by the prospect or customer.

Simple pricing usually wins. Nobody wants to learn your fancy pricing model–not even enterprise procurement teams (even though this is, basically, their job). Complexity makes it harder for you to implement, maintain and experiment. Complexity adds barriers to new customers feeling comfortable buying and extends sales cycles, resulting in slower growth. It also raises the likelihood of a customer gaming your pricing in a way you did not foresee. And nobody likes surprises. While you might think it would be great to send a customer a surprise $1 million overage bill, this is not a winning strategy. Because, among other things, it is not considered fair–after all, you invented your complex pricing scheme that led to this overage; they did not want to bother learning it.

Implementing pricing principles with tariffs

Pricing nerds call the way you extract value a “tariff”. There are one-, two-, and three-part tariffs that encompass pretty much every possible way you can charge for SaaS and AI software: