What happened
OpenAI's Help Center article "Understanding ChatGPT and Codex usage limits" says that starting in December, the usage allowance on Plus, Pro, and Business plans will be shared across ChatGPT and Codex. OpenAI ties the change to bringing the agentic capabilities of Work into the main ChatGPT experience. Usage will be measured in tokens, the units of information a model reads and writes. OpenAI says how much of the allowance you use depends on the model you choose and the task you ask for. More advanced models and complex tasks generally use more. Instant chat will not draw from the allowance. Users will be able to see combined remaining usage and reset times in Settings, then Usage. At the limit, OpenAI says you can wait for the reset, upgrade your plan, or buy credits to continue using eligible features. OpenAI also says it expects most users will not wait for a reset more often than they do now. The article gives only "December," with no exact date and no allowance size for each plan.
Why it matters for business owners
Many owners budget AI the way they budget software: a price per seat, times the number of seats. This change is a reminder that for agent-style work, the seat price is only the entry fee. OpenAI's own documentation describes usage as depending on the model, where the task runs, task complexity, context, reasoning, and tools. It says a long-running task can use substantially more than a short request. Two people on the same plan can have very different experiences, and very different effects on their own limits. The second reason is simpler. Chat and coding or agent work used to be separate buckets for many users. A pooled allowance means one heavy use can crowd out another. The person who runs long agent tasks and the person who only wants to draft emails now draw on the same pool, each on their own seat.
What owners should not misunderstand
This is not a price increase announcement. OpenAI has not published a new price, and it says it expects most users to see little change. Do not assume your costs rise, and do not assume they stay flat. The allowance sizes have not been published. It does not affect everyone equally. The article covers Plus, Pro, and Business plans. Enterprise and API usage are billed under different arrangements. OpenAI's Codex documentation says Codex usage on token-based Enterprise agreements is billed in US dollars, and that your admin can tell you which billing model applies. If you are on a different plan, check what applies to you. Instant chat staying out of the allowance does not mean all chat is free of limits. OpenAI says Instant chat will not draw from usage. Chats on more advanced models are part of what the pool covers. A token limit is not a quality measure. A bigger allowance does not mean better results, and a smaller one does not mean worse. It tells you how much work you can run before you hit a wall, which matters most for the tasks that are long, repeated, or left running. Finally, this is one vendor's plan design. Other vendors meter differently. The lesson is about reading the limit, not about OpenAI being unusual.
The operational lesson
Price a seat by the work it will do, not by the label on the plan. For each AI workflow you care about, you want to know three things: how often it runs, how large each run is, and what happens when the limit hits. A research task that runs once a week is a different thing from an agent that works through a backlog every day. A limit that interrupts a person drafting a memo is a nuisance. A limit that stops a step in a customer-facing process is an operational problem. This is also where cost and workflow meet. Some work belongs on a lighter model. Some belongs in an API workflow where you can see and cap spend, instead of a shared seat pool. Some should not be automated yet. You cannot make that call from the plan page. You make it from the workflow.
What a serious business should do next
Find out which plan each person actually uses, and which of them use Codex or agent features at all. Many teams pay for seats nobody has looked at. Separate light users from heavy users. Light chat and drafting is one need. Long agent runs and coding work is another. Do not assume one plan type fits both. When December arrives, look at the Usage page for the real numbers on your own accounts. Check it every week or two for the first month. OpenAI's own pricing page suggests the same habit. Your data will tell you more than any estimate. Decide in advance what happens at the limit. Who may buy credits, who approves it, and what the monthly ceiling is. Otherwise the decision gets made by whoever is annoyed that day. Do not build a critical process on a seat limit you cannot see. If a workflow must run reliably, it needs usage you can measure and a plan for when it stops. If you are not sure yet, keep it as a person-assisted step until you have a month of real data.
The Atlacis view
Atlacis helps owners slow down, understand the workflow, map the risk, and choose the right tool or system before spending. Usage limits are one of the least visible parts of that picture, because they show up after the purchase. If you are trying to work out which AI plans, tools, or setups match how your team actually works, a short conversation can help you see it before you commit to more seats.
The short version
- OpenAI's Help Center says that starting in December, Plus, Pro, and Business plans will share one usage allowance across ChatGPT and Codex.
- Usage will be measured in tokens, and Instant chat will not draw from the allowance.
- OpenAI has published no exact date and no allowance size per plan, and has not announced a price change.
- Usage depends on model, task, context, and tools, so two seats on one plan can behave very differently.
- Decide in advance who may buy extra credits and what the monthly ceiling is.
- Do not run a critical workflow on a limit you cannot see. Measure a month of real usage first.