The GenAI workload you are costing — e.g. 'contract review', 'sales email drafting', 'support ticket summarisation'.

All costs are shown in this currency. Prices convert from USD using indicative (non-live) FX rates.

People who will actually use the tool in scope — not the whole org. Drives total token volume.

Each prompt = one model call. 8/day is a light knowledge worker; 20/day is a power user.

≈0.75 of a word, or ~4 chars. 800 tokens ≈ a 600-word brief + system instructions. Pasting a long doc inflates this fast.

What the model generates. 400 tokens ≈ 300 words. Output is billed ~5× input, so this is the bigger lever.

Business days the workload runs. 20 is standard; use 22 for a 5-day operational team.

List price for the selected tier in USD. Edit if you have a negotiated or committed-use rate.

Usually 3–5× the input price. Edit if you have a negotiated rate.

Providers discount cached repeated prefixes (system prompts, fixed context) — up to 50–80% off input on the cached portion. 0 if you don't cache.

Per-user-per-month seats. Defaults to your active-user count; reduce if only some users need a seat.

Flat enterprise seat price (e.g. ChatGPT Enterprise ~$30–60, Copilot ~$30). You pay this even for idle users.