ChatGPT or Microsoft Copilot for Your Employees?
Microsoft Copilot is strongest when the value lies in your own M365 data – email, documents, meetings – and in central administration. ChatGPT offers freer workflows and fast access to OpenAI's latest models. Both cost roughly a few hundred kronor per user per month. The worst option is offering nothing at all – that's when shadow AI grows.
ChatGPT or Microsoft Copilot is rarely a question of models – under the hood, both run on OpenAI models. It’s a question of where the value should be created: in your existing documents, email, and meetings, or in a freer tool that employees shape themselves. And for the IT lead, it’s also a question of governance – with a third option that’s worst of all: offering nothing.
Two different promises
Microsoft 365 Copilot promises to make your own data actionable: summarize the email thread, draft the memo based on the documents in the Teams channel, recap what was said in the meeting you missed. The value grows with how much of your work life already lives in M365 – and with how clean your permissions are, since Copilot sees exactly what the user has access to.
ChatGPT promises the best general-purpose tool: writing, analyzing, reasoning, coding, building custom GPTs for recurring tasks – with fast access to OpenAI’s latest models and features, often ahead of the rest of the market. The value grows with employees’ own creativity, and the business plans have caught up on administration: SSO, workspaces, and data controls.
Integration versus flexibility
The difference shows up in daily use. Copilot lives inside the tools employees already have open – Word, Outlook, Teams, Excel – and so requires almost no change in behavior. That’s both its strength and its limitation: it’s at its best on M365-shaped tasks.
ChatGPT lives in its own window and requires the employee to actively go there – but in exchange isn’t tied to any document type or workflow. Creative and analytical roles tend to squeeze more value out of that freedom; administratively heavy roles often get more from Copilot’s closeness to the inbox and calendar.
Licenses, administration, and governance
| Decision point | What you're comparing |
|---|---|
| Price | A few hundred kronor per user per month for both – check current list prices per plan |
| Identity and access | SSO and central user management – Copilot via your existing M365 administration, ChatGPT via the business plan's admin console |
| Data protection | Contract terms on training, storage, and retention per plan – request a written statement for your tier |
| Governance | Ability to restrict features, log usage, and set policies per group |
| Permissions | For Copilot: a permissions cleanup in M365 before rollout – the bot sees everything the user sees |
One detail deserves bold text: Copilot rollouts expose old permission sins. Documents that have sat open to the whole organization for years without anyone finding them suddenly become searchable via a single question. Do the permissions review first, not after the first incident.
The risk of not choosing: shadow AI
While the decision process drags on, a significant share of employees are already using AI – with private accounts, on consumer terms, outside all governance. Customer lists get pasted into free services, contract text gets summarized in tools with no agreement in place, and no log exists when the question comes from audit or a regulator.
That’s why “wait and see” is the most expensive option. A sanctioned tool with reasonable terms, SSO, and a clear policy moves usage to where it can be governed. Which tool you pick matters less than the chosen tool being at least as convenient as employees’ private alternative – otherwise the shadow use continues.
How to choose in practice
A scenario: an organization with 300 employees runs a three-month pilot – Copilot for 30 people in administrative and leadership roles, ChatGPT for 30 in analysis, development, and communications. After three months, they measure weekly usage, self-reported time savings, and concrete examples. The outcome is rarely either-or: many land on Copilot broadly for M365-heavy roles and ChatGPT targeted at specialist roles – or the reverse, depending on the business.
And when your needs outgrow off-the-shelf tools – when the AI needs to reach into your own systems and processes – the next step is a custom AI solution with your data on your terms. At Weapp we’re happy to help you see where that line is; get in touch for a no-obligation conversation.
Frequently asked questions
Can Copilot really use our documents and email?
Yes, that's the core of the product: Copilot pulls context from your Microsoft 365 content – email, files, calendars, meetings – within each user's own permissions. That makes clean permissions a precondition: Copilot surfaces everything a user already has access to, including things that should have been locked down.
Are the models trained on our data?
In the business plans from both Microsoft and OpenAI, the baseline is that your data isn't used to train the models, per each vendor's contract terms. The terms do differ by plan and change over time, though – check the current agreement for the exact tier you're signing, and document the statement.
What do the tools cost per user?
Both run roughly a few hundred kronor per user per month in their business plans, with differences between tiers and over time. Also budget for indirect costs: rollout, training, and governance. Check current list prices – they're adjusted regularly and differ between plans.
Does every employee need a license from day one?
No, and it's rarely wise. Start with a pilot group covering different roles, measure actual use and value per role group, and then scale to the groups where the effect is proven. Paid licenses that go unused are the most common hidden cost in AI rollouts.
What is shadow AI and why is it a problem?
Shadow AI is when employees use private AI accounts for work tasks without the company's knowledge. That puts company data outside any contract, governance, or logging – pasted into services under consumer terms. The risk drops sharply once the company offers a sanctioned alternative that's just as convenient.