The Lineage of Accounting Concepts Leading Up to AI Costing
Why Token Costing Feels New
In conversations about the cost of tokens in AI systems, it feels like we are dealing with something completely new. But if we look at the structure of the problem itself, many of the underlying ideas have been with us for a long time.
The same logic that is reflected in Activity Based Costing is also reflected in the way we think about the cost of a unit of work, or a job. In manufacturing and other operational environments, a job is usually a batch of products or a specific order that is produced together because it shares the same requirements, and the method used to understand the cost of that batch is called Job Order Costing. It’s a way of bringing together the costs associated with that batch of work and understanding how those costs connect to the activities required to produce it.
If we follow that lineage forward into the world of AI, we find that a token is essentially another unit of work. It’s smaller and faster and digital instead of physical, but the underlying idea is the same. Something is being produced, the process to produce it consumes resources, and those resources have a cost. The structure of the costing problem hasn’t changed, but the domain has.
What Practitioners Already Know
People who work with ABC and JOC have more of the foundation for token costing than they realize, even if the terminology makes it feel unfamiliar at first. There is a planned cost for producing a unit of work and a maximum number of units that can be produced. There’s also an actual cost that reflects what really happened and an actual number of units produced. The difference between the two scenarios tells us something about how efficiently the system operated during a period of time or related to a particular job. In manufacturing, this is called a capacity variance. In AI it’s conceptually the difference between what we expected a token to cost and what it actually cost.
Familiar Patterns in a New Domain
The ideas around token economics may sound new, and that’s part of what makes this moment so exciting. But when we slow down and look at the shape of the work, we can see patterns we’ve used for a long time, and there’s something reassuring about that. It gives us a familiar starting point as we move into the next part of the conversation.
