How credit-based pricing works for AI tools
Per-seat pricing does not fit tools where the real cost is compute. Here is how credit-based pricing works, where it helps, and what to watch for.
Most software is priced per seat because the cost of adding a user is close to zero. AI tools are different. Every draft, score, or search runs a model that costs real money, and that cost varies with how much you use it. Per-seat pricing hides that, which is why a lot of AI products have moved to credits.
Here is how the model works and what to look for.
What a credit actually is
A credit is a unit of usage. Each action in the product costs a set number of credits: generating a draft might cost a few, running a discovery scan a fraction of one, calling an external data provider more. Your plan includes a monthly grant of credits, and you can buy more if you run out.
The point of credits is to connect what you pay to what the work costs. A light user and a heavy user pay differently because they consume differently, even on the same plan.
Why not just charge per seat
Per-seat works when usage per user is roughly predictable. With AI it is not. One user might generate ten drafts a month, another a thousand. If both pay the same seat price, either the light user is overpaying or the heavy user is unprofitable, and usually both.
Flat unlimited plans have the opposite problem. To survive heavy users, the vendor has to set the price high enough to cover the worst case, so everyone pays for the heaviest user in the room. Credits avoid both traps by charging for the work that was actually done.
What good credit pricing looks like
Credits get a bad reputation when they are used to obscure cost. Done well, they do the opposite. A few things to look for:
- A visible cost per action. You should be able to see what each operation costs before you run it, not discover it after.
- A live balance. You should always know how many credits you have left, the same way you can see a fuel gauge.
- A clear path when you run out. Hitting the cap should offer an obvious choice: upgrade the plan or buy a topup. It should not silently fail or quietly degrade quality.
- Non-AI features that keep working. Running out of credits should pause the metered AI work, not lock you out of your own data.
If a credit system hides the per-action cost or makes the balance hard to find, that is a sign it is being used to make spend feel smaller than it is.
What to watch for as a buyer
Credits are honest only if the conversion is honest. Two questions worth asking:
- What does a typical month of real usage cost in credits? Ask for an example, not just the per-action rate.
- Do unused credits roll over, or expire at the end of the month? Both are common, but you should know which.
The goal is predictability. You want to be able to look at your plan and your usage and know roughly what next month costs, without a surprise invoice.
How Orviora handles it
Orviora prices AI work in credits because the alternative is either overcharging light users or capping quality for heavy ones. Every action shows its credit cost up front, the Credits dashboard breaks spend down by action, and hitting the cap gives you a clear choice to upgrade or top up. Non-AI features keep working regardless.
You can see the full action-cost map on the Credits page, and the plans on Pricing.