A SaaS Pricing Rethink (Edition 19)
Pricing is undergoing a significant rethink with the onset of AI. In this post, I suggest a few pricing approaches that put more power in customer's hands. And more money in their wallets.
Subscription based pricing was a significant innovation when compared to outright purchasing. For businesses, it gave them revenue predictability. For customers, it allowed them to not just spread their costs over time, but also control (and cut) costs in case of non-use.
While subscription pricing is often associated with SaaS and digital goods, it has been around for a lot longer. One of the earliest known instances was the introduction of a fire insurance scheme by King Charles I of England in the mid-1600s. News and magazine publishers followed suit closely in the late 1600s.
However, subscription pricing for physical vs. virtual goods should be seen distinctly. For physical goods, such as magazines, it often simply represents confirmed repeat purchases. For virtual goods, especially those with no (or miniscule) running costs, it can be a lot more about having a control of costs, and paying mostly for what you use. Or need.
Instead of paying $100 for a software license, you pay $5 per month, with an option to stop at the end of every 12 months. While you will end up paying more if you end up using that software for more than 20 months, you are paying a much smaller amount upfront. Plus you can stop at the end of 12 months, at which point you would have paid only $60.
You therefore have a lot more control.
In the SaaS software world, this has been the dominant pricing model for 25+ years.
The case for change
I believe this needs to change. In fact, the onset of AI is already bringing new models to the fore - including consumption-based pricing, outcomes-based pricing; and various combinations thereof.
The reason why I think it needs to change is simple. Too many software licenses go unused these days. Which means the customer is paying for something that they are not utilizing. Pricing needs to start with ‘customer value’. Anything that does not deliver value to customers should be rethought.
The onset of a new paradigm in AI, as well as new delivery models like ‘Service as a Software’ gives us an opportunity to rethink pricing in a fundamental manner. I believe we should jump at the opportunity. It would be especially easier for new companies. It also gives them an opportunity to use pricing as a lever to unseat incumbents.
Here are the 3 kinds of pricing strategies that should be considered.
Tiered License-based pricing
Existing subscription aka license-based pricing has its own advantages (predictability, for one). I believe it should continue. However, companies should start offering (and customers should demand) pricing tiers - within pricing plans.
For example, let us say you have 100 salespeople, but only 30 of them consistently use Salesforce. In that case, rather than you having to buy 100 Salesforce licenses, you should be able to buy 30 licenses. In case 3 of the other 70 salespeople decide to use Salesforce in a given month, you should have the flexibility to pay for additional three licenses for that given month only.
You will very likely pay more per license if you are purchasing 30 licenses instead of 100, but the odds are that overall, you will end up saving significantly.
Consumption-based pricing
There are scenarios where it is easy to measure the utilization of a particular product or service. Often, the utilization in such cases can be broken into structured units/parts. Cloud, streaming services or lately, AI agents offer such possibilities.
Activity-based pricing, where an AI agent might handle a ticket or handle a customer query, would usually fall under this bucket too.
The problem with consumption-based pricing is that the CFO does not have full control over costing and it could play havoc with budgets. One way to handle this concern could be to have upper limits. In other words, you pay per activity/unit consumed, but once it crosses a certain limit, you could only be billed a certain pre-agreed amount. This amount would likely be 20-30% higher than the amount you might have paid for an equivalent license with unlimited activities, but it allows you to manage your costs for non-use significantly.
Upper limits might not be possible for services that have high marginal input costs, but for most software where the marginal costs are negligible, such upper limits should be possible.
At Humantic AI, we are considering such pricing for our new Account Research agent Miia (along with standard license based pricing).
Outcomes-based pricing
The final major pricing model that is coming into play with AI is outcomes-based pricing. In this model, customers don’t pay for a license or an activity, they pay for an outcome delivered.
For example, if a customer pays a vendor for a ticket that an AI handles, then that would be consumption-based pricing. However, if the customer pays for a ticket that the AI resolves, that would be outcomes-based pricing.
Similarly, for Miia, our Account Research agent, if a customer pays per Research report generated, it would be consumption-based pricing. However, if a customer pays for the increase in revenue based on that particular report, it would be outcomes-based pricing.
It’s probably apparent that for outcomes-based pricing to become an option, it should be possible to clearly measure - and attribute - the impact of the given product or service on the given outcome. If such an impact cannot be measured or clearly attributed, outcomes-based pricing would very likely not succeed.
In the field of Sales, such pricing is a lot more possible than some other fields, as Sales is an outcomes-driven field. I have written more about outcomes-based pricing for Sales in an earlier post.
One of the often-discussed pricing model for AI agents is to price them in terms of people that they replace. Y Combinator has argued that this can be 10X bigger. Their argument seems to be that AI agent providers can then tap into manpower budgets, which tend to be 10X larger than tool budgets.
While seemingly lucrative, I don’t think it is likely to become the dominant pricing model for AI agents. One, pricing will eventually steer towards value/cost or demand/supply, as some people have argued. Two, not all agents replace humans.
Conclusion
Customer-centricity should be at the heart of everything that companies do. That should apply to pricing. We should therefore rethink pricing from the customer’s lens - why should a customer pay more than the corresponding value that they will be capturing?
The primary drivers of pricing should therefore be value and cost. Value should act as the upper bound (one should not charge more than a % of the value they are delivering) and cost should act as the lower bound (one could not charge less than the cost they are accruing).
In fact, one could argue that demand and supply should play a much smaller role in pricing - they become an easy excuse for price gouging otherwise.
At Humantic AI, we teach something called Buyer-First Selling - a way of selling where salespeople learn to look at everything from their prospects’ perspective. And try to solve problems that they need solved, the way they want them solved.
I believe that the same concept should apply to pricing as well.