TL;DR
- Flat saas fees are great but may not scale
- Usage based pricing scales well but customers get spooked
- Outcomes based pricing is the holy grail but is hard to measure
- Best case you should lean towards outcomes based pricing, even if it’s only 70% accurate
Pricing is hard as an AI company. Back when we were building "ChatGPT for customer support", we had a customer paying us $20/m that was very unprofitable. And we also had customers paying us $X00/month that was very profitable.
Here’s a pricing guide based on everything we learned selling an AI product.
For context, we had (and still have) small customers from $20/m to big customers paying us thousands of dollars a month.
What are my pricing options?- Flat fee pricing- Pure usage based pricing- Outcomes based pricing
Flat fee pricing (e.g. $200/month no matter the usage) is the easiest to understand. It’s also what your seed to series B companies probably prefer too.
Why?
We found that most of these companies were very scared of usage based pricing because they had no good way to estimate and “cap” their spending. Almost everything in the $100-500/m range was better off with a flat fee pricing.
If you need to win deals and your customers are concerned of crazy bills, this is your best bet.
In fact, we found that we could even charge a small premium and customers would still close because of “peace of mind”. But the biggest problem is you don’t scale as your customer scales. We’ve had to renegotiate pricing multiple times with customers because of this (annoying for all parties).
Pure usage based pricing is billing someone based on everything they use, regardless of outcomes. For example, OpenAI bills you for all the tokens you use, even if their responses are bad.
This one’s a little hard to sell, especially to executives. They almost never care (or understand) what tokens/requests/messages are. They only care about outcomes. And they get nervous because they don’t understand how much “usage” is needed to get the outcomes they want.
For example, if you sell an “AI outbound cold email tool” to a VP of sales, they’re immediately thinking “Does it work? If it does, will this scale? How much usage will I need to hit my results?”. The harder it is to calculate this, the less likely they are to convert. (We have seen other companies do some fancy calculators to help calculate, no idea how effective it is though.)
Outcomes based pricing is the holy grail. This is like “$1 per lead closed by AI” or “$1 per support ticket resolved”. The hard part is (1) measurement and (2) attribution.
For example, Intercom Fin charges per ticket resolution. But how do you measure that? If a customer doesn’t respond, does that mean their problem is solved? Or they were frustrated because the answer sucks and they left? (Intercom Fin has really high churn from business customers because they bill a non-response as a resolution).
Long term, I’d push a lot of AI companies to try to go this route. This is the holy grail of AI and aligns incentives between you and your customer. Yes, it’s not perfect. Yes, it’s hard to measure/attribute. But I think it makes a lot of sense.
So what do I recommend?
It depends.
If you’re selling to small startups and need to close deals fast, flat fee pricing is the easiest to understand and advocate for.
If you’re selling an outcome that you can easily measure and attribute, outcomes based pricing all the way. Even if you’re roughly able to measure and attribute (e.g. with 70% accuracy), you might consider going this route because then you and your customer’s incentives are aligned. We’ve found if you get them results, they’re happy to have some flexibility on accuracy.
If you’re selling to developers or a customer who already understands usage based pricing (e.g. users of Twilio/AWS/etc), then you might have some luck with usage based pricing. For non-developers or people higher up, we generally found usage based pricing didn’t clearly communicate our value, which is important.
Also, generally, if the usage is estimated to be less than $100/m, I’d recommend going with flat fee pricing with a cap. Easier to understand and makes your revenue more predictable (small customers can have volatile usage and high churn).
Need help pricing? Or tracking your usage and billing?
I’m happy to help. We’re building Spryngtime, an SDK and platform to track AI usage & billing (free to sign up and track 10,000 events per month). Use us instead of building usage analytics & billing yourself (like what we did
)!
If you need help designing your AI product pricing to close early customers, feel free to DM or email me and I’d be happy to hop on a call!
- Flat saas fees are great but may not scale
- Usage based pricing scales well but customers get spooked
- Outcomes based pricing is the holy grail but is hard to measure
- Best case you should lean towards outcomes based pricing, even if it’s only 70% accurate
Pricing is hard as an AI company. Back when we were building "ChatGPT for customer support", we had a customer paying us $20/m that was very unprofitable. And we also had customers paying us $X00/month that was very profitable.
Here’s a pricing guide based on everything we learned selling an AI product.
For context, we had (and still have) small customers from $20/m to big customers paying us thousands of dollars a month.
What are my pricing options?- Flat fee pricing- Pure usage based pricing- Outcomes based pricing
Flat fee pricing (e.g. $200/month no matter the usage) is the easiest to understand. It’s also what your seed to series B companies probably prefer too.
Why?
We found that most of these companies were very scared of usage based pricing because they had no good way to estimate and “cap” their spending. Almost everything in the $100-500/m range was better off with a flat fee pricing.
If you need to win deals and your customers are concerned of crazy bills, this is your best bet.
In fact, we found that we could even charge a small premium and customers would still close because of “peace of mind”. But the biggest problem is you don’t scale as your customer scales. We’ve had to renegotiate pricing multiple times with customers because of this (annoying for all parties).
Pure usage based pricing is billing someone based on everything they use, regardless of outcomes. For example, OpenAI bills you for all the tokens you use, even if their responses are bad.
This one’s a little hard to sell, especially to executives. They almost never care (or understand) what tokens/requests/messages are. They only care about outcomes. And they get nervous because they don’t understand how much “usage” is needed to get the outcomes they want.
For example, if you sell an “AI outbound cold email tool” to a VP of sales, they’re immediately thinking “Does it work? If it does, will this scale? How much usage will I need to hit my results?”. The harder it is to calculate this, the less likely they are to convert. (We have seen other companies do some fancy calculators to help calculate, no idea how effective it is though.)
Outcomes based pricing is the holy grail. This is like “$1 per lead closed by AI” or “$1 per support ticket resolved”. The hard part is (1) measurement and (2) attribution.
For example, Intercom Fin charges per ticket resolution. But how do you measure that? If a customer doesn’t respond, does that mean their problem is solved? Or they were frustrated because the answer sucks and they left? (Intercom Fin has really high churn from business customers because they bill a non-response as a resolution).
Long term, I’d push a lot of AI companies to try to go this route. This is the holy grail of AI and aligns incentives between you and your customer. Yes, it’s not perfect. Yes, it’s hard to measure/attribute. But I think it makes a lot of sense.
So what do I recommend?
It depends.
If you’re selling to small startups and need to close deals fast, flat fee pricing is the easiest to understand and advocate for.
If you’re selling an outcome that you can easily measure and attribute, outcomes based pricing all the way. Even if you’re roughly able to measure and attribute (e.g. with 70% accuracy), you might consider going this route because then you and your customer’s incentives are aligned. We’ve found if you get them results, they’re happy to have some flexibility on accuracy.
If you’re selling to developers or a customer who already understands usage based pricing (e.g. users of Twilio/AWS/etc), then you might have some luck with usage based pricing. For non-developers or people higher up, we generally found usage based pricing didn’t clearly communicate our value, which is important.
Also, generally, if the usage is estimated to be less than $100/m, I’d recommend going with flat fee pricing with a cap. Easier to understand and makes your revenue more predictable (small customers can have volatile usage and high churn).
Need help pricing? Or tracking your usage and billing?
I’m happy to help. We’re building Spryngtime, an SDK and platform to track AI usage & billing (free to sign up and track 10,000 events per month). Use us instead of building usage analytics & billing yourself (like what we did
)!If you need help designing your AI product pricing to close early customers, feel free to DM or email me and I’d be happy to hop on a call!