It is stated that Microsoft is evaluating new options to reduce costs in corporate artificial intelligence services. It is reported that the company has made the Copilot Cowork service for Microsoft 365 Copilot users generally available and has also changed its pricing model. With this change, a system based on usage amount will be implemented instead of a fixed fee approach. Thus, companies will pay depending on how much they use artificial intelligence tools. This step by Microsoft aims to make the high transaction costs created by artificial intelligence agents taking on increasingly complex tasks more sustainable.
Recently, enterprise artificial intelligence agents can not only carry out simple commands, but also manage multi-stage business processes. Operations such as classifying e-mails, meeting preparations, collecting information from Teams conversations or creating reports from Excel data can be carried out in a single workflow. However, such task chains require significant processing power. While each new step means using more computing resources, costs can rise rapidly in large-scale corporate uses.
In the new pricing model announced by Microsoft, companies will be charged based on usage units called “Copilot Credits”. This system aims to more accurately reflect the background processing load of the tasks being performed. Thus, it is aimed to create a more balanced cost structure between businesses with limited use and organizations with intense artificial intelligence workload.
Microsoft examines DeepSeek V4 and open source alternatives
It is stated that Microsoft is working on different model options in order to further reduce costs. It is stated that the company is evaluating customized versions of DeepSeek V4 or similar open source models, in addition to the currently used OpenAI and Anthropic models. DeepSeek has recently become one of the notable names in the artificial intelligence industry. Models that stand out especially in terms of cost-performance balance can provide results close to the performance offered by more advanced and higher-cost models in some scenarios.
Despite this, data security and compliance requirements are critical for corporate customers. For this reason, it is stated that a possible DeepSeek integration will work entirely on Azure infrastructure. Thus, customer data can be processed without leaving Microsoft’s cloud environment. In addition, companies’ existing controls regarding data placement, legal compliance and security policies will be preserved. Especially for organizations working with sensitive data, this approach can reduce questions about the use of outsourced models.
One of the recent trends in the artificial intelligence market is that businesses focus not only on performance but also on total cost of ownership. While model capabilities were at the forefront in the first generation of productive artificial intelligence applications, today many institutions evaluate usage costs more carefully. As a result, open source models and self-hosted solutions are gaining more traction.
The model diversity approach that Microsoft is working on can also be considered a reflection of this change. Offering options at different cost and performance levels according to the needs of companies can make it easier for artificial intelligence technologies to reach a wider user base. In the coming period, it will become clearer which model Microsoft will choose and how corporate customers will respond to this new approach. The current picture is that more emphasis is placed on cost and flexibility issues so that advanced artificial intelligence agents can be used more widely in daily business processes.