ChatGPT Business vs a company brain: when $25 a seat is enough

ChatGPT Business is enough for a small business when the job is helping each person write and research faster, using knowledge already in its connected tools. It costs $25 per user a month on monthly billing, per OpenAI's help center. A company brain earns its price when the problem is knowledge in people's heads and repeat work nobody owns.

You'll hear "just use the $20 plan" when a founder asks about paying for AI help, and a version of it is in our own FAQ: can't we build this with ChatGPT? It deserves a straight answer, because for a lot of teams ChatGPT on its own is the right choice. This piece covers what ChatGPT Business costs and does today, when it's all you need, and where a company brain picks up.

What ChatGPT Business costs today

ChatGPT Business costs $25 per user a month for a Standard seat on monthly billing, or $20 per user a month if you pay for the year, according to OpenAI's ChatGPT Business help article. The "$20 plan" could mean that annual price or a personal plan, and OpenAI's pricing FAQ says Free, Go and Plus are designed for individuals.

Plan	Price per user	What changes	Source
Business, Standard seat	$25 a month billed monthly, $20 a month billed annually	All ChatGPT features, including Work and Codex, with included usage limits	OpenAI help center
Business, Premium seat	$125 a month billed monthly, $100 a month billed annually	5x the usage of a Standard seat, no five-hour limit and access to dots, which OpenAI calls always-on agents that proactively move work forward	OpenAI business pricing, ChatGPT Business page
Enterprise	Contact sales	SCIM, role-based access controls, data residency, custom retention	OpenAI business pricing

A few terms from the same help articles matter for a small company. A workspace needs at least 2 paid seats, and OpenAI's General FAQ caps a Business subscription at 200 paid seats for workspaces created since August 24, 2026. API usage is billed separately from the seats. ChatGPT Team and ChatGPT Business are the same plan: OpenAI renamed Team to Business on August 29, 2025, per its help center.

As an illustration, for a 20-person company on monthly billing, the math is 20 seats x $25 x 12 months = $6,000 a year, or $4,800 a year on annual billing, using OpenAI's listed Standard prices. That's a small number next to almost any salary, which is why the objection lands.

What you get for $25 a seat

A Business seat now does far more than chat. OpenAI has been pushing hard at small companies: in July it launched a ChatGPT for small businesses program with webinars, in-person training and partner plugins, per Inc. and OpenAI's own announcement. The plan itself includes several things that used to require a custom build:

• Company Knowledge. It searches your connected apps and only shows each person what they could already open in the source tool, per OpenAI's Company Knowledge help article.
• Workspace agents. You can build an agent for a repeatable task, connect it to apps, share it with the team, run it on a schedule and use it in Slack, per OpenAI's workspace agents help article.
• ChatGPT Work. An agent that takes a goal and works through a multi-step task, per the ChatGPT Business page.
• Privacy terms. OpenAI's business pricing page lists no training on your business data by default. Our piece on AI and company data covers what that means in practice.

That's a lot of product for the price, and for many small teams it covers everything they'd ask of AI right now.

When ChatGPT Business is enough

ChatGPT Business is enough when most of the value you want is one person getting help with their own work. If these sound like your company, buy the seats:

• The main need is drafting and research: emails, proposals, content, summaries of long documents, first passes at spreadsheets.
• Your knowledge is already written down. The SOPs are current, decisions get recorded in Slack or a doc, and the source of truth lives in tools OpenAI connects to.
• Your stack is mainstream: Google Workspace or Microsoft 365 plus Slack, nothing homegrown.
• Someone on the team likes this work and has a few hours a week to build agents and fix them when a connection breaks.
• You're fine with your company's AI history living inside OpenAI's workspace.

If that's you, $25 a seat is a good deal, and we'd say the same on an intro call. Buying a company brain now would mean paying for plumbing you don't need yet.

Where a seat stops

A ChatGPT seat stops where the knowledge was never written down and where nobody owns the agents. Those are the cases where teams come to us.

It can only search what exists in writing. Company Knowledge looks across your connected tools when someone asks a question. If the reason you dropped a supplier was said out loud in an unrecorded meeting and lives in your ops lead's head, there's nothing for it to find, and when she leaves, it walks out the door with her. Business seats do include record mode in the macOS desktop app, which transcribes and summarizes meetings and can auto-join Google Meet calls from a connected calendar, per OpenAI's ChatGPT Record help article. Those notes land in each user's own history, and there's still no shared running log of decisions and their reasoning. A company brain is built around feeds: every meeting and customer conversation goes into one memory as it happens, and decisions get written down with the reasoning behind them.

Somebody has to own the agents. Workspace agents are a real step forward, and they still need a person to design each one and keep its connections running. OpenAI's own help article tells you to connect Slack agents through a service account and to scope access down, because a personal connection can expose your data to anyone who uses the agent. That setup is fine for a small number of agents. Premium seats add dots, which OpenAI's Business page describes as always-on agents that proactively move work forward, and someone still sets their priorities and checks their work. For a sales follow-up after every call, a briefing before every meeting, a weekly scoreboard and invoice chasing, it becomes a part-time job, and it's the work that quietly stops when the person who set it up gets busy.

Some tools have no standard connection. OpenAI's Company Knowledge article says it can reach custom apps built with MCP, a protocol for connecting AI to outside tools. Someone on your side has to write and maintain that connector for your industry software or the database your ops team lives in.

You can't take it with you. Per OpenAI's General FAQ, self-service data export isn't available in Business workspaces, and a canceled workspace may be deactivated, leaving members without access. In the builds we do, the memory and the agents run on machines the client controls, and the client owns all of it.

The bill grows with headcount. Seats are priced per user, so every hire adds a line. Hockeystick Team is a flat $199 a month for unlimited people, so a new hire costs nothing extra.

ChatGPT Business and a company brain, side by side

	ChatGPT Business	Hockeystick
Price	$25 per user a month, or $20 billed annually (OpenAI)	$199 a month flat for your whole team, unlimited people ($49 for one person). A custom build starts at $15,000
Where knowledge comes from	Searches connected apps at the moment you ask, plus per-user meeting transcripts from record mode	Feeds that capture every meeting and customer conversation as it happens
The record of decisions	Whatever someone wrote into a connected tool	Decisions written down with the reasoning, each answer showing its source
Agents	Workspace agents your team builds and maintains	An agent that works from the brain, with a library of ready-made skills; we build custom skills on a company build
Tools with no standard connector	You write a custom MCP connector	Request it on a hosted plan, or have it built as part of a company build
If you leave	No self-service export; deactivated workspaces lose access	Cancel any time; a custom build runs on machines you control
Best for	Individual drafting and research on well-documented tools	Knowledge in people's heads, repeat work across the team

Running both makes sense for many teams. Seats give everyone strong help with writing and research, and the brain holds the record and runs the repeat work underneath.

The math on each side

A ChatGPT seat is priced per user, so its cost is the seat count times the price. On monthly billing, 8 Standard seats cost 8 x $25 = $200 a month, about the same as Hockeystick Team's flat $199. From there every hire widens the gap: a 20-person team pays 20 x $25 = $500 a month for seats, against the same $199. The time the work saves needs its own payback calculation, and we'd rather you run it than take our word for it.

Our Leverage Audit prices time as hours a week x hourly rate x 52 weeks, using set hourly rates by role, such as $75 for sales and $150 for a founder. Say your sales team spends 6 hours a week writing follow-ups after calls: 6 x $75 x 52 = $23,400 a year. Say your founder spends 2 hours a week preparing for meetings: 2 x $150 x 52 = $15,600 a year. Add 8 hours a week of support replies at the audit's $45 support rate: 8 x $45 x 52 = $18,720 a year. Those three workflows come to $57,720 a year of staff time, and that's the figure to hold any quote against, ours included.

Those figures are an example, and your hours will differ. Per the audit's live engine, a 16 to 50 person team lands around $300,000 a year across its top 3 workflows, counting staff time only, so treat it as a floor. A ChatGPT seat can recover some of that time for the person using it. An agent drafts the follow-up after every call, so nobody has to think to ask.

Hockeystick costs $49 or $199 a month, so you can check those numbers against a real month of use. If you want it built for you, a custom build starts at $15,000. We price that after a free strategy call, so bring numbers like these to the call and hold our quote against them.

What it looks like running

Rye Valley, the holding company that runs our own install, removed $425,724 in annual expenses within the first month after install, and the team shipped 17x as many commits in the 75 days after as in the 75 days before. Its brain keeps a running record of what was decided and why, which is the part a chat seat doesn't produce on its own.

How to decide

Start with the seats if you haven't pushed them yet. Turn on Company Knowledge, build one workspace agent for your most annoying weekly task, and see whether it sticks after a month. If it does, you may never need us. If it stalls because the knowledge you need was never written down or because nobody has time to keep the agents running, a build starts to earn its price. Our builds start with a free 30-minute strategy call and a fixed quote with a written definition of done, and the brain and first agents go live in the first month, usually sooner. The cost guide breaks down what moves a quote.