If your company is building agents, it is highly likely you will end up monetizing with credits.
Credits allow you to align value with price and cost. And compared to pure pay-as-you-go models, they typically have built-in predictability for both buyers and sellers.
Ok, makes sense…
But what are product and engineering teams having to do to figure out their approach to managing credits?
Often the investigation begins on Claude, where the founder, or a product manager, or a billing engineer might ask "how should I handle credit pricing for this product" (and share an md file)?
Then Claude reads what companies like Schematic , Stripe , and others have written on the subject, and comes back with thoughts and answers.
And from there, companies usually have to start thinking about a few things:
How will we count credit usage?
How will we manage a credit ledger?
How do we enforce our monetization policies (e.g. what happens next when a user burns through their credit grant?)
This third question contains a lot of complexity… and it's easy to see why…
Do we want to turn off access?
Do we want auto top-ups?
Do we want the customer to have to manually top up?
Do we want overage fees?
What if the user hasn't used all their credits, but the period has ended… do we want credits to expire or roll over?
Which is the best decision for revenue quality? What's best for revenue recognition? For product adoption? For driving net dollar retention?
And how will we actually enforce any of this in the product?
More questions start emerging about the tools we'll need...
Can we do all of this in our existing billing system or do we now need to buy a metering platform? If so, what's the cost? What are the offerings out there?
So now we're trying to figure out our business systems infrastructure…
Maybe we're starting to think, okay, our billing system can do the counting and the credit ledger… but oh wait, can it do the real-time enforcement…?
I don't think so... how will we handle that? Do we also now need an entitlements platform too? Or do we build that?
If we build an entitlements platform, are we kind of just rebuilding a metering platform… so should we just build it all ourselves?
What does the entitlements platform need to do? How will we ensure extremely low latency and high reliability?
Ok, this feels like it's going to be a lot of work.
And then a new set of questions…
There's a ton of UI complexity and requirements here to sell credit-based pricing to enterprises so CFOs and their procurement teams will trust our pricing.
How will we surface usage to users?
Will we make it visible by user type, by humans and by agents?
Do we want our users to be able to forecast their usage, and potentially their expansion? Do we want to give our customers some degree of control?
What if the user wants to see their usage?
How will we surface all of this? What if the user wants to control their own usage within the org? Org-level controls… user-level controls… role-level controls… actor-level controls (agents vs. humans).
Quickly, companies realize that the move to AI and to consumption and to credits is a big iceberg of an initiative with a ton of edges.
And whatever they start with will change quickly, either because they'll get it wrong, or because they're selling to enterprise and they'll need to support custom plans.
Selling AI means a few big things:
We are a new company with a new business model.
Change is really hard, so we need great culture.
As much as great culture, we need iteration speed and business agility. Our mindset has to be the learner's mindset.
And as we move to credits and usage, we have one outstanding requirement: establish trust with our customer at all times. They need the same things we need. Transparency, control, and visibility.
If you've found yourself working through this stuff, ping us at Schematic.
We pride ourselves on our ability to partner with companies at this inflection and help them cross the chasm successfully.
