Enterprise AI agents are becoming highly effective at handling complex, multi-step tasks like sorting emails, preparing meetings, or generating reports across Outlook, Teams, and Excel.
However, this capability comes with significant computational costs, leading to rising bills for businesses.
Microsoft is stepping in to make these tools more practical for everyday use.
The company has made Copilot Cowork generally available to Microsoft 365 Copilot customers and shifted to usage-based pricing.
Instead of a flat fee, companies now pay based on actual "Copilot Credits," which meter the computational load for each task.
Exploring Cost-Effective Models
Beyond the pricing shift, Microsoft is exploring ways to further reduce costs.
Reports indicate the company is considering a fine-tuned, self-hosted version of DeepSeek V4 or another open-source model as a more affordable option alongside existing Anthropic and OpenAI models.
DeepSeek V4 has demonstrated strong performance at much lower prices, sometimes costing a fraction of frontier models per token.
Any integration would be optional and run entirely on Azure, ensuring customer data remains within Microsoft's secure cloud environment with enterprise-grade protections, compliance, and data residency controls.
This approach addresses potential concerns about using a Chinese-developed model in sensitive business settings.
The move reflects the maturing AI industry, where agentic AI's ability to maintain context through long task chains makes it expensive to run at scale.
By offering model choices and metered billing, Microsoft aims to strike a balance that enables broader adoption without sticker shock.
Observers will watch which lower-cost model the company selects and how customers respond in the coming weeks.
