Quick answer
Copilot is not one product, which is why its pricing confuses people. There is a free chat assistant, a consumer subscription, a per-user add-on for Microsoft 365 business plans, Copilot inside GitHub, and consumption-billed agent capability in Copilot Studio. The one that costs real money is the Microsoft 365 add-on, billed per user per month on top of an existing licence — and the trap is that it is frequently sold with an annual commitment for every seat when only some of those people will use it.
On figures: Microsoft revises Copilot packaging, naming and pricing frequently, and it varies by region, currency and agreement type. This guide explains the structure, which is stable, and the traps, which are consistent. For current prices check Microsoft’s own pricing pages or your reseller — not any third-party article, including this one.
The reason there are six-thousand-word articles about Copilot pricing is that Microsoft attached one brand to several unrelated products sold on different models. Untangling that is most of the work.
| Product | What it is | Billing model | Who it is for |
|---|---|---|---|
| Copilot (free) | Web and app chat assistant | Free | Anyone. General chat use |
| Copilot Pro | Consumer subscription with app integration | Per person, monthly | Individuals using personal Microsoft 365 |
| Microsoft 365 Copilot | Embedded in Word, Excel, Outlook, Teams, with access to your organisation’s data | Per user, per month, on top of an eligible M365 licence | Businesses. This is the expensive one |
| GitHub Copilot | Coding assistance | Per user, per month, separate tiers | Developers. Entirely separate purchase |
| Copilot Studio | Build custom agents and copilots | Consumption-based, message packs or capacity | Teams building internal tools |
| Security Copilot | Security operations assistance | Consumption-based compute units | Security teams |
The critical distinction is between rows one to two and row three. Free Copilot and Copilot Pro give you a capable assistant. Microsoft 365 Copilot gives you an assistant that can see your organisation’s documents, emails and meetings — which is the actual product people are buying, and why it costs what it does.
Microsoft 365 Copilot is an add-on, not a standalone product. That has consequences.
Watch out
Copilot respects existing Microsoft 365 permissions — it surfaces what a user could already find. That sounds reassuring and is the opposite in most organisations, because years of over-permissive SharePoint sites and shared folders mean people can technically access far more than anyone realised. Copilot makes that latent access usable: a salary spreadsheet nobody could find in a forgotten site is now one natural-language question away. Audit permissions before deployment, not after the first incident.
My take
Buy a small number of seats for people whose work is genuinely document- and meeting-heavy, run it for a quarter, and measure. Not satisfaction surveys — actual usage data, which Microsoft’s admin reporting will give you. Most organisations discover a clear split: a group for whom it is transformative and a majority who forgot it exists. Buying for that majority upfront is how a sensible tool becomes a bad investment.
Excel is the consistent disappointment relative to expectations. People assume it will do sophisticated analysis; it works best on clean, well-structured tables and struggles with the merged cells and inconsistent layouts that real spreadsheets contain.
| Situation | Verdict |
|---|---|
| Heavy Microsoft 365 user, lots of meetings and documents | Likely yes, for those specific people |
| Small team, data not in Microsoft 365 | No. You lose the part you are paying for |
| Mostly need drafting and general chat | A general model subscription is cheaper and as good |
| Developers | GitHub Copilot, which is a different and clearer purchase |
| Permissions never audited | Fix that first. Deployment will expose it |
| Want custom internal agents | Copilot Studio, and model the consumption billing carefully |
Microsoft 365 Copilot is a genuinely good product sold in a way that encourages overbuying. Its distinctive value — grounding in your own organisation’s content — is real and unavailable from a general chatbot at any price. It is also worthless if your content is not there, or if your permissions are a mess.
Pilot narrowly, measure real usage rather than enthusiasm, audit permissions before rather than after, and budget for the data work nobody puts in the proposal. Do that and the economics usually work for a meaningful subset of your staff.
If what you actually want is a good assistant for drafting and thinking, buy a general model subscription instead. It costs less and does that job just as well.
For how Copilot compares with the alternatives in daily use, see Copilot vs ChatGPT vs Gemini. For the underlying model behaviour, what ChatGPT is and how it works.
It depends which Copilot. There is a free chat assistant, a consumer Copilot Pro subscription, Microsoft 365 Copilot billed per user per month on top of an existing eligible licence, GitHub Copilot sold separately for developers, and consumption-billed Copilot Studio and Security Copilot. Check Microsoft’s own pricing pages, as packaging and prices change frequently and vary by region.
No, it is an add-on requiring a qualifying Microsoft 365 licence underneath it. The real cost per head is both figures combined, which business cases routinely miss. It also depends on your data actually being in Microsoft 365 — organisations storing documents on file servers or in Google Drive get a small fraction of the value.
It respects existing Microsoft 365 permissions, surfacing only what a user could already access. In practice that is a risk rather than a reassurance, because years of over-permissive SharePoint sites mean people can often reach far more than anyone realised. Copilot makes that latent access usable through natural-language search, so audit permissions before deploying.
Committing annually to seats for everyone. Adoption is consistently uneven — some staff use it daily, many try it twice — so buying organisation-wide on a twelve-month term means paying a year for substantial non-use. Start with a subset of document- and meeting-heavy roles and expand based on measured usage.
No, they are entirely separate purchases with separate pricing tiers. Buying Microsoft 365 Copilot gives your developers nothing for their coding work, and buying GitHub Copilot gives your wider staff nothing in Word, Outlook or Teams. Budget them as two distinct products.
Meeting recap in Teams is the most consistently valued: summaries, decisions, actions, and asking what was said about a topic in a meeting you missed. Outlook thread catch-up and drafting come second, since high-frequency tasks compound small savings. Finding documents across a large organisation without knowing where they live is also genuinely strong.
Because expectations are set by demos using clean, well-structured tables. Real spreadsheets contain merged cells, inconsistent layouts, multiple header rows and data entered as text. Copilot works well on tidy data and struggles with the messy reality, so the gap between demo and daily experience is wider in Excel than in any other app.
If you mainly want drafting, thinking and general chat, a general model subscription costs less and does that job equally well. Copilot’s distinctive value is grounding in your own organisation’s documents, emails and meetings, which no general chatbot offers at any price — but that value only exists if your content is genuinely in Microsoft 365.
A comparison of eight BPM platforms on published licensing, BPMN 2.0 support and architecture, plus…
A practical Triumphoid guide to rankmath content ai vs external ai tools: why i do…
Hard metrics on memory overhead, concurrent executions, and CPU optimization at scale.
Perplexity searches, reads and cites. How retrieval-augmented answers work, why citations are not verification, and…
A practical Triumphoid guide to trello vs airtable for wordpress editorial automation: why simpler usually…
Pillar covering token-cost and infrastructure-optimization techniques for CTOs and BizOps leaders facing scaling API bills.