What ChatGPT Business is
ChatGPT Business is OpenAI's self-serve plan for teams, formerly called ChatGPT Team. It gives an organization a shared workspace with centralized billing and admin controls. OpenAI states that workspace data, including conversations and memories, is not used to train its models.
As of November 2026, OpenAI's help center lists Standard seats at $25 per user per month on monthly billing, or $20 per user per month billed annually, with a workspace requiring at least two paid seats. A higher Premium seat tier is also listed. Prices are in US dollars and may vary by country.
For a five-person team that is a small, predictable cost for a very capable assistant that every member can use immediately.
What a custom agent is
A custom agent is software we build around your team's work. It uses the same class of models, but it runs on a schedule, keeps a memory you control, connects to your tools and follows rules you set.
The policy team agent we built is a concrete example. It lives in WhatsApp. Each morning it logs new publications and events in the team's field, and team members can get a list of their open items. It keeps shared logs of decisions and document status, drafts documents in the team's own format, profiles experts, translates documents, and drafts messages to teammates that a person confirms before they are sent.
None of that happens in a chat window unless somebody opens it and asks.
Side by side
| ChatGPT Business | Custom AI agent | |
|---|---|---|
| Cost | $20 to $25 per user per month, minimum two seats (as of November 2026) | Project build plus a retainer to run and improve it, quoted by PI after a free call |
| Time to start | Minutes | Weeks, after mapping the work |
| Where it works | ChatGPT apps and web, plus connectors | WhatsApp, Slack, email, Google Workspace, your website or internal tools |
| Scheduled work | You start each conversation | Runs on a schedule: morning digests, daily open-item lists, monitoring |
| Shared memory | Workspace features and memory per OpenAI's settings | Shared logs of decisions, roles, deadlines and style that apply to the whole team |
| Guardrails | Admin controls over the workspace | Rules about what it may send, to whom, and what needs human approval |
| Maintenance | None, OpenAI runs it | PI reviews and improves it weekly |
| Best for | General assistance for every person on the team | One or two workflows that must run reliably without being asked |
Where ChatGPT Business is the better choice
If the need is general, ChatGPT Business wins on every practical measure. It costs little, starts immediately, and gives each person a capable assistant for writing, analysis, research and coding. The admin controls and the training exclusion answer the main concerns a small company has about using AI at work.
It is also the right first step even if you think you need a custom agent. Using it for a month shows you which tasks the team keeps repeating, and those repeated tasks are what a custom agent should automate.
Where a custom agent is the better choice
- The work has to happen without being asked: a morning digest, a daily reminder, monitoring a website or inbox.
- The team needs one shared memory, so agreed decisions apply across the team, and the assistant should know who is who.
- The assistant must work where the team already is, such as WhatsApp or email, rather than in a separate app.
- There are rules about sending. Messages to colleagues or partners should be drafted, confirmed and then sent, with a person approving each one.
- Output must follow house templates every time, not just when the prompt is written carefully.
A sensible path for a five-person team
Start with ChatGPT Business for everyone. After a few weeks, list the tasks people keep doing by hand that a chat window cannot take over: the scheduled checks, the shared tracking, the messages that need drafting in a specific format.
Then build one agent for that list. Keep its scope narrow, give it clear rules about what it may send, and review what it does each week. That is how we built the policy team agent, and it is the approach we recommend to anyone who asks.