Usage-based pricing for work management

Seat pricing forces a binary: someone is either worth a full licence or excluded. Real operations don't work that way - they include reviewers, approvers, and external collaborators who touch work occasionally. Paying full seats for all of them is how the bill balloons without matching value.

Usage-based pricing separates the two costs that actually differ: lightweight participation (which should be cheap or free) and heavy AI execution (which has a real compute cost). Charging for consumption instead of headcount aligns the bill with the work.

A token is the unit AI models read and write in - roughly a word-piece. When an agent reads a task, reasons about it, and writes an update, it consumes input and output tokens. Metering in tokens means you pay for the actual model work done, not a flat fee that over- or under-charges depending on use.

The fear with usage pricing is a runaway bill. The answer is admin-side controls: budgets per team, caps that stop spend before it happens, and visibility into where tokens go. Predictability comes from governance, not from a flat fee.

• Set budgets per team or workspace so no group can overspend silently.
• Use hard caps for environments where overage is unacceptable.
• Review usage by workflow to find the automations worth keeping or trimming.

Two models exist for paying the AI cost. Bundled means the platform marks up and resells model usage. Bring-your-own-key (BYOK) means you connect your own model-provider account and pay the provider directly - useful when you already have committed spend, want a specific region, or need the AI cost on your own books.

BYOK also reinforces sovereignty: the model endpoint and keys are yours, so AI execution can stay in your region and under your contracts. The right choice depends on whether you value simplicity (bundled) or control and direct accounting (BYOK).