Flat pricing vs per seat for AI tools: the math

Where per-seat pricing came from

Per seat was a fair proxy for a long time. When software is a tool a person operates, like a spreadsheet or a design app, each seat roughly equals one person getting value, so charging by seat tracked value delivered. It also made budgets easy: headcount times price.

The model survives less because it still fits and more because billing systems, sales compensation, and revenue forecasting are all built on it. That's not a scandal, it's inertia. But it means the pricing model most AI tools inherited was designed for a different kind of software.

What AI changes

An AI teammate isn't a tool each person operates, it's closer to shared staff. When it assembles the Monday digest or drafts the follow-up emails, that work happens once and the whole company gets it. The value scales with the work done, not with how many people can log in.

Per-seat pricing on top of that creates a strange incentive: every additional person who might benefit is a billing decision. Teams respond exactly as you'd expect, sharing logins and leaving out the people who'd use it least, which quietly caps how much of the company the AI actually helps.

The math at small-team sizes

Take an illustrative per-seat price of $25 a month, made up for the arithmetic rather than quoted from anyone. At 4 people that's $100 a month, at 15 it's $375, at 40 it's $1,000. The bill triples and then triples again while the AI's actual work, the digests and drafts and watchers, may barely change.

Glitch Team Zone's tiers are flat: Starter is $99 a month and Growth is $299, whatever your headcount does. The crossover happens fast. On the illustrative numbers, flat wins from around a dozen people on the low tier, and the gap only widens as you hire. More to the point, growth stops being a software-budget event.

The guest problem

Small businesses don't stop at their org chart. Clients, vendors, bookkeepers, and contractors all belong in the conversation sometimes, and per-seat pricing makes every one of them a purchasing decision. So they get left out, and the coordination the tool was supposed to hold moves back into email.

Flat pricing removes that decision entirely. In Glitch Team Zone, outside people join guest channels without touching the bill, which means the question is only ever whether someone belongs in the conversation, never whether they're worth a seat.

Where usage costs fit

There's one honest complication: AI has a real marginal cost. Tokens cost money, so a flat price with truly unlimited AI would be quietly subsidized by light users or quietly degraded for heavy ones. A pricing model that pretends otherwise is hiding the meter somewhere.

The structure we think is fair: a flat tier that includes a monthly AI allowance, and past it, usage passes through at provider price with no markup, under a budget cap an admin sets. No margin on tokens means no incentive for the vendor to want you burning more of them. And if you bring your own LLM key, your usage runs on your own provider account with no limit from us at all.

When per seat is still fine

If value really does arrive one person at a time, per seat remains a fair proxy. A design tool a single person lives in all day, or a development environment, prices per person reasonably, because the person is doing the work and the tool is amplifying that one person.

The test for any AI tool is one question: does the value scale with the number of people logged in, or with the work the AI does? If it's the work, a per-seat price is measuring the wrong thing, and you'll feel it every time you hire.

Common questions