Impact Pricing Blog

I keep disagreeing with my friend Steven Forth about credits. 

I’ve learned that when I disagree with Steven, it means I don’t understand the issue deeply enough yet. Or one of us is missing a layer. Sometimes it’s him. Often it’s me. Almost always, the truth sits underneath the disagreement.

Our debate is about whether credits are a pricing model or a billing system. The answer, annoyingly, is both.

My instinct has always been to say credits are billing. If a company prices on a single metric like tokens, API calls, or minutes of compute, then selling credits in advance is just prepayment. Buyers are paying early for known consumption. Nothing about value changes. Nothing about willingness to pay changes. It is billing, simply made more convenient.

Steven treats credits as a pricing model, and at first I pushed back. Pricing models answer the question, “What do buyers pay for?” Credits, on their own, don’t do that. They don’t explain outcomes. They don’t explain why the product is worth buying. They only show up after a buyer has already decided to buy.

The disagreement forced me to go one layer deeper.

Credits change character the moment they span more than one metric.

As soon as a company lets credits apply across multiple forms of consumption, tokens, agent runs, tool calls, storage, latency guarantees, credits stop being neutral billing. These things are not naturally comparable. When a company creates a single pool of credits and applies them across different activities, it is setting exchange rates between fundamentally different kinds of value consumption.

That is currency creation.

Currencies do work. They force tradeoffs. They shape behavior. They embed assumptions about equivalence. One agent run equals this many tokens. One minute of compute equals that many tool calls. Those ratios are pricing decisions, whether the company acknowledges them or not.

The cell phone analogy makes this concrete.

Early mobile plans sold minutes. Those minutes were prepaid usage credits. It was one metric. Pure billing abstraction.

Then texting arrived. At first, texts were priced separately because carriers did not know how people would value them. Later came text bundles. Still separate meters. Minutes here. Texts there.

Now imagine carriers had instead said: You buy 1,000 credits per month. A voice minute costs 2 credits. A text costs 1 credit.

That would not have been billing convenience. It would have been a pricing decision. The carrier would be declaring that one minute of voice time is worth exactly two texts. That ratio would shape behavior and lock the carrier into defending it. When texting exploded, the carrier would face a trap. Change the rate and anger customers, or keep it stable and watch margins erode.

Eventually, carriers abandoned the currency entirely and sold predictability instead. Unlimited plans were not generosity. They were an admission that minutes and texts were no longer aligned with what buyers valued.

That same arc is playing out in AI.

Credits start as billing. When they span metrics, they become currency. That feels like pricing.

But they are still not a pricing strategy.

Credits do not explain why someone should buy. They only operate after value has already been assumed. Mistaking currency for value is the real risk.

Disagreeing with Steven is almost always insightful. In this case, credits have a dual role. Confusion starts when we pretend they only have one.  We are using one word for two purposes: billing and currency.  Maybe we need a new word.  

Oh, and here is a recent article so you can read Steven’s thoughts directly.

Share your comments on the LinkedIn post.

Now, go make an impact!