Which AI Plan a One-Person Business Actually Needs

Two different AI vendors, checked on the same afternoon, quote a small business the same four numbers.

A standard seat is $20 per user per month billed annually, $25 billed monthly. A premium seat is $100 per user per month annually, $125 monthly. The premium seat buys five times the usage of the standard one. And the workspace will not open at all for fewer than two seats.

That is OpenAI’s ChatGPT Business and Anthropic’s Claude Team, read on September 14, 2026. Two companies, no coordination, an identical ladder down to the dollar.

When two competitors converge that precisely, the pricing is telling you what is actually being sold. It is not features. It is throughput and administration.

The ladder is priced on usage, not capability

Walk any of these pricing pages and look for the thing the cheap tier cannot do.

You will mostly not find one. The same models answer on the $20 plan and the $100 plan. What changes is how much you may ask before the meter stops you, and, on the business tiers, whether an administrator can see and control the account.

Claude’s own page says the premium seat gives “5x more usage than standard seats.” OpenAI’s says its premium seat is worth five times a standard one and that a standard seat carries a five-hour usage limit. Both describe a ceiling, not a locked door.

This reframes the question. “Which plan do I need” is not a features question you can answer by reading a comparison table. It is a measurement question: how often do you hit the ceiling, and what does hitting it cost you? If you are still deciding what you would even point an assistant at, the seven jobs an agent can take is the better place to start than a pricing page.

Most owners have never measured that, which is why they answer it by buying up.

The two-seat minimum is the trap for a business of one

Both vendors require a minimum of two paid seats before a business workspace exists.

For a one-person business that is a hard floor of $40 a month, annually committed, to buy administrative controls over an organization of one — user provisioning, shared workspace policy, an audit trail of your own behavior. A sole operator is the administrator. There is nothing to administer.

So the tier labeled Business is, for many actual small businesses, the wrong tier. The individual plans are not a lesser product; they are the product without the org chart.

I run this operation on a single individual power-tier subscription. No seats, no workspace, no admin console. It is the largest line on my software bill by a wide margin and it is still an individual plan, because the second seat would have bought me a permissions model and nothing else.

There is a real threshold here and it is not revenue or ambition. It is the first moment somebody other than you needs their own login. Before that, seats are overhead.

Check what you already bought

Before pricing a new subscription, look at the suite you are already paying for.

Google’s business plans now carry the assistant inside the apps, with the entry plan described as limited access and the middle and upper plans as the full feature set (Google Workspace plans). The same pattern holds across the major productivity suites. A business paying for email and file storage may already own an assistant good enough for drafting, summarizing, and search, and be shopping for a second one out of habit.

That is not an argument for the bundled tool. Bundled assistants are usually weaker at the long, messy, multi-step work, which is exactly the work worth paying for. It is an argument for knowing what is already on the bill before adding to it. I found $280 of subscriptions in my own books last week that had never generated a charge, which is the same failure in the other direction: nobody reconciles a small recurring line.

The three lines that multiply the bill after month three

The advertised number is not what most businesses end up paying. Three mechanisms do the work, and all three are quiet.

The seat you added and never removed. Per-seat plans bill on commitment, not on use. A contractor added in month two is a full year of that seat. Nothing in the interface will remind you; the charge is stable, which is precisely why it becomes invisible. Every phantom line I found in my own books had been stable for months.

Usage that spills past the ceiling. Once a workspace turns on flexible pricing, usage beyond the included limits draws against credits — OpenAI’s documentation describes exactly this. The subscription stops being a fixed cost and becomes a variable one, and a variable cost with no line-item statement is a cost nobody audits. The tier was the budget. Credits remove the budget.

The annual discount you did not actually take. The headline price on every ladder above is the annually committed one. Monthly billing is 25 percent higher on every tier at both vendors — $25 against $20, $125 against $100. Businesses that start monthly to stay flexible are paying a quarter more, indefinitely, for an option they rarely exercise.

None of these is a trick. They are ordinary subscription mechanics that happen to compound when the product is priced by seat and metered by usage at the same time.

How to actually pick

Three counts, each of which you can do in an afternoon, none of which requires a forecast.

Count the days last month you hit a limit. Zero means you are on the right tier or a tier too high. If it happened often enough that you rearranged your work around it, the five-times step is cheaper than the rearranging — that is the entire case for the premium tier, and it is a measurable one.

Count the people who need their own login. One means you do not need a business plan, and the two-seat minimum makes it strictly worse than the individual tier. Two or more with real separation of duties means you do.

Count what the narrow build would replace. If a tier upgrade is what lets you stop renting a packaged tool, price both sides. My own first-party ledger replaced software that runs $456 to $900 a year at the tier that handles a rental property, and it took an afternoon — the full bill and the projects it paid for that never shipped are itemized here. The upgrade is worth what the thing it displaces is worth, not what the feature list implies.

What you should not do is pick the tier by reading the comparison grid. The grid is designed to make the top row look necessary, and on a ladder priced by throughput it mostly is not.

The commitment

Here is the falsifiable version.

As of September 2026, the major assistant tiers are separated by usage allowance and administrative control, not by capability. If a vendor tells you the higher tier unlocks something the lower one cannot do at all, open the pricing page and find that thing in writing. Sometimes it exists — longer outputs, earlier access to new features, compliance controls a regulated business genuinely needs. Usually the difference is a multiplier.

I expect this to hold for a while and I expect it to eventually stop holding, because pricing converges until somebody breaks it. When a tier starts gating a capability rather than a quantity, the advice above stops working and the comparison grid starts mattering again.

Until then, the honest answer to “which plan do I need” is that nobody can tell you from the outside, including me. It depends on a number only you can measure, and most owners have not measured it.


More in this series at The Owner’s Desk. Prices quoted were read from the vendors’ own pages on 2026-09-14 and will change. The argument is about how the tiers are structured, which changes more slowly. No vendor named here is a client, a partner, or an affiliate.

Written by Colin Reed

Colin Reed is an operations executive and independent consultant whose career spans foodservice distribution, airline catering, last-mile delivery, and hospitality, from restaurants and bars to luxury resort food and beverage programs. He has held distribution P&L with capital authority and run enterprise system selections across warehouse management, labor management, and ERP platforms, and advises funds, corporates, and management consultancies through expert networks and research firms. On Modern Wealth Model he writes on applied AI in the physical economy, Bitcoin, and long-horizon retirement planning — and builds the calculators.

Full bio · Contact · Consulting