Short answer: from November 2026, Microsoft is putting usage limits on more Copilot Chat features for staff who don't have a paid Copilot licence, and it won't say what the limits are. If someone on your team has quietly built a daily task around the free, bundled AI, that task can now stall in the middle of a working day. Spend an hour this week finding out which work depends on it.
What Microsoft changed
Copilot Chat comes with Microsoft 365 business subscriptions. Some of its features, such as image generation, already had limits for people without a paid Microsoft 365 Copilot licence. Microsoft's partner announcement dated 4 November 2026 says it is now adding "fair use" limits to more Copilot Chat features for those users.
The dates depend on company size. The rollout began on 2 November 2026 for organisations with 2,000 or fewer Microsoft 365 seats, so if your company is that size, it has already started. Larger organisations follow from 9 November.
A few details matter for anyone running a team:
- Nobody on your side can see the limit. Microsoft sets and manages the limits. Admins can't view or change them, and Microsoft doesn't publish the numbers because they're subject to change.
- The limit can move during the day. The announcement says fair use access "can vary throughout the day".
- Work in progress is safe. A reply that's already being written isn't cut off. The next request is the one that waits.
- The message comes with an upgrade button. In most organisations, the limit notice lets the user request a paid licence from their admin, and where the company allows it, buy one directly.
Microsoft's support page on Copilot Chat limits puts it plainly for users: hitting a limit "isn't an error or product malfunction." Copilot tells you in the product when a feature isn't available.
Why this catches small teams out
In a lot of small companies, nobody decided to use AI. People started using what was already on their screen. An accounts executive summarises vendor emails in Copilot every morning. A sales coordinator drafts follow-ups in it. Someone in HR rewrites job posts with it. None of this is written down, so nobody knows it's there until it stops working.
The included version also does less than the paid one, which is easy to miss. Microsoft's own comparison of Copilot Chat with and without a licence lists standard access to AI models against priority access, and work data limited to files you upload against full access to meetings, emails, chats and files. So the free version may already be doing a smaller job than people assume.
This isn't only a Microsoft question. Google's help page for Gemini on personal Google accounts describes compute-based limits that depend on how complex your prompt is, which model and features you use, and how long the chat runs, refreshing every five hours until you reach a weekly limit. If your staff use personal AI accounts for company work, they're living with limits you can't see either, and that's a separate data question as well.
A one-hour check for this week
You don't need a project for this. One person, one spreadsheet, one hour.
1. Ask each team lead one question. "Which tasks would slow down tomorrow if Copilot, ChatGPT or Gemini stopped working for a day?" Write down the task, who does it, how often, and which tool and account they use (work or personal).
2. Mark the ones that run every day. A weekly job post rewrite can wait a few hours. A morning summary of 40 supplier emails that feeds the purchase team can't.
3. Check which account each daily task runs on. In Microsoft, the label under a user's name in Copilot, such as Copilot Chat (Basic) or M365 Copilot (Premium), tells you which version they have. Anything running on a personal account goes on a separate list for your AI use policy.
4. Decide per person, not per company. For the few people whose daily work stops without AI, a paid licence or a properly built workflow is probably worth it. For everyone else, the included version with limits is fine. Buying licences for the whole company because one person hit a cap is how AI spend creeps up. Our one-hour audit of AI tool spend covers the other side of this.
5. Tell your help desk or office manager. Microsoft's own advice to admins is to brief help desk staff and review licence request settings. In a 30-person company that's usually one person. Let them know a limit message is expected behaviour, and decide who approves licence requests before the first one lands.
The bigger point
Free and bundled AI is a good way to start, and most teams should keep using it. The risk is building real, recurring work on a version whose terms the vendor can change without telling you the numbers. That's fine for drafting an email. It's not fine for the morning report your operations head reads at 10am.
The fix is the same as for any other tool your business relies on: know what depends on it, know who owns it, and pay for the parts that matter. Do it now, while a limit is still a small annoyance and not a missed dispatch.
Arthat AI
Working on something similar at your company?
Arthat AI builds these automations for Indian businesses every week. If you want to talk through whether this fits your situation, we'll give you a straight answer — even if the answer is "don't build it."
