This guide explains both models in plain language: how each one works, who it punishes and who it rewards, what happens with a real team of 8, and when per-seat pricing genuinely makes sense.
Per-seat: you pay for people, not for work
Per-seat pricing is simple to state: every person with access costs a fixed monthly fee, whether they live in the tool or never open it. Simplicity is its only virtue. In practice it has three effects that rarely get said out loud.
- It punishes growth: hire 3 people and your bill goes up automatically, even if the amount of work being managed stays the same. Growing the team becomes a software budgeting decision.
- It punishes occasional users: the accountant who logs in twice a month to check a status costs exactly as much as the PM who spends all day in the tool. Many teams end up sharing accounts or leaving people out just to save money.
- In the AI era it often comes with double billing: the seat on one line, AI credits on another. You pay to be there, then pay again to use the part that actually saves you time.
Usage-based: you pay for what the AI executes
Usage-based pricing (pay as you go) flips the logic: you don't pay for how many people log in, you pay for how much work the AI does. That work is measured in tokens — a token is, roughly, a small chunk of text (a short word or part of a longer one) that the AI reads or writes while handling your request. Ask the AI to build a schedule, draft a summary, or follow up on tasks, and it consumes tokens; if nobody uses the AI one month, you pay almost nothing that month.
The most important consequence: inviting people is free. Your accountant, your client, the freelance designer collaborating for one month — all of them can log in, see the board, and comment without adding a cent to the bill. Cost follows value: you pay when the AI works, not when someone looks. That's how Doku works: a free plan with no credit card to start, and on the paid plan AI is billed per token, with zero per-seat cost.
The pattern that's getting old
Look at the pricing pages of the most heavily advertised AI project management tools — Motion is the classic example — and a pattern repeats: an expensive per-person monthly seat, AI credits billed separately when you want the smart features, and a trial that demands a credit card before you can even try it. Every piece of the pattern pushes in the same direction: commit before you know whether the tool works for you, and pay for installed capacity instead of work done. It isn't a scam — it's the subscription-software model from a decade ago, applied to a product whose real cost no longer depends on headcount but on AI usage.
A team of 8 where 3 people use the tool daily
Put people numbers (not dollar figures) on a typical case: an 8-person company where the PM, the coordinator, and the ops lead live in the tool, two people check in a couple of times a week, and the other three — accounting, a partner, an external contractor — peek in a couple of times a month.
With per-seat pricing
You pay for 8 full seats, even though 5 people barely use the tool. The alternative is giving access to only 3 and leaving everyone else asking for screenshots over chat — which is exactly the problem the tool was supposed to solve.
With usage-based pricing
All 8 people get in for free. You pay for the tokens the 3 heavy users consume when the AI builds plans, summarizes progress, or follows up. The 5 occasional users see everything, comment, and cost nothing.
When you hire 3 more people
Per-seat: the bill goes up automatically, whether they use the tool or not. Usage-based: the bill only moves if those people actually put the AI to work. Growing the team stops being a software decision.
When a quiet month comes along
Per-seat: you pay the same as in your busiest month of the year. Usage-based: consumption drops on its own, because the AI worked less. The bill breathes with your business.
When per-seat pricing does make sense
Let's be honest: per-seat isn't inherently bad — it's bad when usage is uneven. If every single person on your team uses the tool intensively every day (think a support team inside its helpdesk, or a sales team inside its CRM), a flat per-person fee gives you a predictable cost and can genuinely compete with paying per action. The problem is that project management almost never looks like that: most teams have 2 or 3 daily users and a ring of occasional collaborators. There, the seat overcharges you — every single month.
Token transparency: what drives your spend and how to see it
The reasonable fear about pay as you go is the end-of-month surprise. The answer is transparency, and it helps to know what actually moves consumption:
- Long, context-heavy requests consume more than short ones: asking the AI to summarize a six-month project costs more tokens than asking it to create a task.
- Automatic features (smart reminders, follow-ups, recurring summaries) consume little per event, but they add up if you enable them across many projects at once.
- Creating, moving, and completing tasks by hand consumes no AI at all — you only pay when the AI reads or writes on your behalf.
- In Doku you can check your accumulated spend for the period at any time, so the bill is never a black box; and on Enterprise, a prepaid token vault lets you fix the budget upfront with unlimited users.
Try usage-based pricing — no credit card
Start free with unlimited tasks and the basic AI assistant. When you want more AI, you pay only for the tokens it consumes — inviting your team costs nothing, whether it's 3 people or 30.
Start free with DokuFrequently asked questions
What is a token in AI pricing?
A token is the smallest unit of text an AI model reads or writes: roughly a short word, or a fragment of a longer one. When you ask the AI for something, both your request and its response are measured in tokens, and usage-based pricing bills exactly that amount. Think of it as the kilowatt-hour of AI: a standard measure of work performed.
How do I know how much I'll spend with usage-based pricing?
Start free and watch your real consumption for the first few weeks — serious tools, Doku included, show your accumulated spend for the period at all times. After two or three weeks of normal use you'll have a realistic monthly projection based on your actual team, not on a vendor's estimate.
What happens if I go over budget?
With usage-based pricing your spend is visible in real time, so a spike gets caught in days, not when the invoice lands. You can dial back automatic features or pause AI on projects that don't need it. For hard budget caps, Doku's Enterprise plan uses a prepaid token vault: you load a fixed amount and the AI draws from it, making surprises impossible.
Why do most tools charge per seat?
Because it's the model inherited from subscription software: easy for the vendor to forecast and comfortable to bill, since revenue stays flat even when usage drops. Before AI it had some logic — the vendor's cost barely varied with usage. With AI, real cost does depend on usage, and charging a seat plus separate credits passes that difference to the customer twice.
Is usage-based pricing always cheaper than per-seat?
Not always, and be wary of anyone who says otherwise. If your entire team uses the tool intensively every day, a flat per-person fee can be competitive. Usage-based pricing wins when usage is uneven — a few heavy users and several occasional ones — which is how the vast majority of project teams actually work.