Microsoft just made one of the more interesting moves in enterprise AI. They're bringing Anthropic's Claude into Microsoft 365 Copilot, alongside their existing OpenAI integration. If you're running a Finnish business on M365 (and statistically, you probably are), this affects you.
Let me explain what actually changes, what stays the same, and why this matters beyond the headline.
What Happened, Specifically
Microsoft announced that M365 Copilot will support multiple AI models, starting with Anthropic's Claude. Previously, Copilot ran exclusively on OpenAI's GPT models. Now you'll have access to Claude's reasoning capabilities directly inside Word, Excel, Outlook, PowerPoint, and Teams.
This isn't a small technical detail. Claude and GPT have meaningfully different strengths. Claude tends to be better at nuanced writing, following complex instructions, and handling long documents. GPT tends to be faster and better at certain structured tasks. Having both means Copilot can route different tasks to whichever model handles them best.
What Gets Better
Document Work in Word
Claude is exceptionally good at working with long, complex documents. Think contracts, reports, regulatory filings. Where GPT sometimes loses track of context in a 40-page document, Claude handles it with more consistency. If your team spends time summarizing long documents or rewriting sections while maintaining tone, this is a real improvement.
Email Drafting in Outlook
This one matters for Finnish businesses specifically. Claude's instruction-following tends to produce more natural-sounding emails across languages. (We explore this further in our article on Claude Cowork for business users.) If you've ever gotten a Copilot-drafted email that sounded oddly American, you know the problem. Claude won't magically write perfect Finnish, but it handles multilingual contexts with more care.
Data Analysis in Excel
Both models can write formulas and analyze data. The real win here is redundancy. If one model struggles with your specific query, Copilot can fall back to the other. Microsoft is building a routing layer that sends each task to the best-suited model automatically.
Meeting Notes in Teams
Claude's summarization capabilities are strong, particularly for meetings where multiple topics overlap. Expect more structured, accurate meeting summaries with clearer action items.
Competition Drives Prices Down
This is the part most people aren't talking about. When Microsoft depended solely on OpenAI, they had limited leverage on pricing. Now they have alternatives. And Anthropic, Google (with Gemini), and open-source models are all competing aggressively.
Microsoft 365 Copilot currently costs $30 per user per month. That's steep for a 20-person company ($600/month, $7,200/year). But competition between AI providers will push that number down over the next 12-18 months. We're already seeing enterprise volume discounts drop below $22 per user.
For Finnish SMBs who've been waiting for Copilot to make financial sense: that moment is approaching.
The Bigger Trend: AI Becomes Infrastructure
Here's what I think matters most about this announcement. It signals that AI is moving from "special project" to "background utility." Like electricity or internet, AI is becoming something that just runs inside your existing tools.
Two years ago, using AI meant going to ChatGPT, copying your text in, getting a response, copying it back. That was clunky. One year ago, Copilot put AI inside Office but it felt bolted on. Now, with multiple models competing to power the same features, AI is becoming genuinely invisible infrastructure.
This changes how you should think about AI adoption:
- Stop treating AI as a separate initiative. It's a feature of your existing tools now. Train your team to use Copilot the same way you trained them to use keyboard shortcuts.
- Budget for it like software, not consulting. AI costs are becoming predictable monthly expenses, not project-based investments.
- Focus on workflows, not models. Which specific model powers your email drafting matters less than whether your email workflow is designed well.
What This Doesn't Solve
I want to be honest about the limits. M365 Copilot, even with Claude, handles individual tasks inside individual apps. It writes an email. It summarizes a document. It creates a chart.
What it doesn't do is connect your systems. It won't automatically take a customer inquiry from your website, check their order history in your ERP, draft a response, and update your CRM. That's workflow automation, and it requires purpose-built solutions.
Think of it this way: Copilot makes individual employees 15-20% more productive at desk work. Custom AI automation can eliminate entire manual processes, saving 5-10 hours per week per workflow.
Both matter. They're just different tools for different problems.
What Finnish Businesses Should Do Right Now
If you're already paying for Copilot: You'll get Claude access automatically as Microsoft rolls it out. No action needed. Watch for the routing features that let you specify which model handles which tasks.
If you've been on the fence about Copilot: This makes the case stronger. Start a pilot with 3-5 power users in your team. Measure actual time savings over 30 days before rolling out company-wide.
If you need more than document-level AI: Copilot handles the surface. For deeper automation (connecting systems, processing data across platforms, building AI agents that work autonomously), you need something custom. That's where the real ROI lives for most businesses.
The AI landscape is moving fast. The companies that win aren't the ones using the fanciest models. They're the ones who figure out which specific problems to solve first and build from there. For a deeper look at how models compare in practice, read our GPT-5 business impact analysis.