What is AI credits (usage metering)?
AI credits are a usage currency that software products use to meter and price AI features. Instead of passing on raw model token costs, the product gives each AI activity, such as drafting a lesson or answering a question, a price in credits. Customers buy or receive credits, spend them as people use features, and track usage against budgets.
How Personal Academy handles it
Personal Academy gives every AI activity a published credit price; AI stops when credits or a person’s weekly allowance run out, and budget alerts fire at 75, 90 and 100 percent.
A worked example
Suppose a 300-person consulting firm turns on AI course drafting and an AI tutor. An administrator checks the credit price for each activity, sets a weekly allowance per person so a few heavy users cannot use the whole balance, and sets a monthly budget. When usage reaches the first alert level, the administrator sees which teams and activities used the most credits and decides whether to buy more.
Large language models are billed by their providers in tokens, roughly pieces of words, with different prices for input and output and for each model. Those costs are hard for customers to predict or understand. Credits translate them into a simpler unit tied to what people actually do in the product.
A credit system has a few parts. A price list assigns each AI activity a credit cost, which may vary with the size of the job or the model used. A balance holds the credits an organisation has received with its plan or bought separately. Metering records each use against the balance, with the person, activity and time. Controls decide what happens when limits are reached.
Good controls prevent surprises. Common ones are per-person allowances, which stop a single user from consuming a shared balance, and budget alerts at set percentages of a monthly amount. A hard stop when credits run out protects against unexpected cost but can interrupt work, so administrators need warnings in time to act.
Transparency matters. Buyers should be able to see the credit price of each activity before using it, how credits relate to money on the published price list, whether unused credits expire, and how usage is reported by person, role and activity. Without that, credits can hide the real cost of AI features.
Credits also let vendors change underlying models without changing customer prices for each activity, while still covering their own costs. When comparing products, ask for the credit price of the activities you expect to use most, and estimate a typical month from them.