
News
Microsoft ditches OpenAI for Anthropic to run Copilot
Disruption snapshot
Microsoft is integrating Anthropic’s Claude into Microsoft 365 Copilot. AI agents will complete workplace tasks end-to-end. Copilot becomes the central interface businesses use to run AI workflows.
Winners: Microsoft and AI infrastructure platforms controlling enterprise distribution. Losers: smaller SaaS tools that automate narrow workflows like reporting, dashboards, or simple internal apps.
Watch Copilot adoption inside Microsoft 365. If usage exceeds 15% of paid seats, enterprise software economics could shift toward AI agents replacing traditional SaaS interfaces.
Microsoft (MSFT) has a Disruption Score of 3.
As if Claude recently becoming one of the fastest growing AI apps in the world wasn't enough.
Anthropic just scored another big win.
Microsoft (MSFT) says it's ditching OpenAI and will use Anthropic instead for its AI Microsoft Copilot.
Microsoft says the new feature, called Copilot Cowork, will run on Anthropic’s Claude Cowork system.
The goal is to let AI handle real workplace tasks from start to finish. Think building internal apps, generating spreadsheets, organizing large datasets, and coordinating projects across teams.
Businesses will also be able to access Anthropic’s latest Claude Sonnet models through Microsoft 365 Copilot, which currently costs $30 per user each month.
At first glance, this looks like another AI feature announcement. But for investors, the bigger story is control.
Microsoft is positioning Copilot as the central hub where businesses interact with AI. Microsoft will control the distribution layer that sits on top of enterprise software, and that’s where the long term value could be created.
The disruption behind the news: Microsoft just connected AI agents to the largest corporate software network in the world.
Most companies testing AI agents run into the same problem. Security teams get nervous when autonomous systems start accessing internal data, databases, and workflows.
Microsoft solves that by embedding agents inside its existing enterprise stack. Microsoft Azure identity. Microsoft 365 permissions. Corporate governance tools that companies already use.
Microsoft already sells productivity software to more than 400 million paid Office users worldwide. If even 10 percent of those seats turn on AI agents at $30 per month, that equals roughly $14B in annual recurring revenue. That estimate doesn’t include additional usage fees for heavier automation workloads.
AI agents act like a new type of middleware for software tasks. They generate spreadsheets instead of requiring users to open Excel templates. They assemble simple applications instead of forcing companies to buy niche SaaS tools. They organize data instead of requiring specialized analytics dashboards.
Much of this shift is happening because AI systems have improved dramatically in capability over the past year, allowing them to handle complex workflows that previously required dedicated software tools.
If Microsoft controls the agent interface, it can gradually absorb entire categories of software.
That’s why investors panicked in February and sold software stocks when agent capabilities improved quickly. Thousands of smaller SaaS tools exist only because humans needed structured interfaces to complete repetitive work. AI agents remove that requirement.
The irony is that Anthropic may have built the agent technology, but Microsoft owns the entry point into the enterprise. Distribution wins again.
There’s also a second layer to the economics. At $30 per user per month, Copilot already costs about 50% more than the typical Microsoft 365 seat. Many enterprise plans average around $18 to $20 per user each month. That flips the historical pricing structure of productivity software. The assistant becomes more expensive than the underlying tools.
If agents start handling even a few hours of administrative work per employee each month, the return on investment becomes obvious for companies. That means Microsoft isn’t just adding another feature. But creating a new, higher margin pricing tier on top of a user base it already controls.
What to watch next
Enterprise AI agents are moving from early demonstrations to real deployment.
The speed of adoption will depend on trust and governance. Microsoft already controls both inside corporate IT.
Watch one number over the next 12 months. Copilot penetration across Microsoft 365 seats. Adoption is still early. If usage climbs past 15 percent of the installed base, the economics of enterprise software start shifting.
Second, watch how fast agent capabilities improve. Right now Copilot Cowork focuses on tasks like spreadsheets, document summaries, and basic app generation. Within 18 to 24 months these systems will likely coordinate multi step workflows across entire departments. Finance automation. Sales reporting. Internal tools. Software maintenance.
That’s when traditional SaaS companies start feeling pressure.
Finally, watch Microsoft’s model strategy. By integrating Anthropic alongside OpenAI, Microsoft is building a model marketplace inside Copilot. That gives enterprises flexibility while tying them more deeply to Microsoft’s infrastructure.
Switching costs increase quickly once workflows depend on embedded AI agents.
This might be the start of Microsoft turning Copilot into the operating system for knowledge work.
Once your AI coworker sits inside the same system that runs your documents, identity, and cloud infrastructure, it gets a lot harder to leave.
Microsoft (MSFT) has a Disruption Score of 3. Click here to learn how we calculate the Disruption Score.
Microsoft is also part of the Disruption Aristocrats, our quarterly list of the world’s top disruptive stocks. P.S: Anthropic is also expanding its enterprise platform capabilities with new features such as free AI memory designed to support long running workflows and attract new customers.
Recommended Articles



