Blog

Microsoft Monetization Model – From Agent Creation to Consumption via Companial

Friday, August 1, 2025
Reading time: 5 minutes

As AI agents become more deeply embedded in Microsoft Business Applications, understanding how they are licensed and consumed is becoming increasingly important. The Microsoft monetisation model is an essential piece of understanding for partners looking to guide customers through this evolving landscape. At its core, Microsoft is introducing a message-based billing system that reflects the actual consumption of AI services.

This article explores the key components of how Microsoft are monetising their Agents and highlights how Companial can support partners in navigating the journey from agent creation to consumption.

Licensing Agents

There are two primary ways to license agent usage, message packs, which provide a fixed number of messages (25,000 per month) for a set annual fee, and pay-as-you-go (PAYG,) which charges per message with billing handled through Azure. PAYG is also eligible for Microsoft Azure Consumption Commitment (MACC) making it an attractive route for enterprise customers in particular, whilst the message packs provide a fixed cost option, up to where the capacity is exhausted on a monthly basis.

The flexibility of the model allows customers to choose the best fit for their needs. Message packs are ideal for predictable usage scenarios, while PAYG offers scalability with limited cost control for variable or high-volume environments. Both models support a wide range of use cases, from customer service to internal productivity agents and can be used across Microsoft 365, Dynamics 365, and Business Central.

Core licensing paths include Business Central (Premium SKU + Copilot Studio Message Pack,) Dynamics 365 (Premium SKUs with Copilot features,) and Microsoft 365 (M365 Copilot users with additional consumption-based billing for agent actions.)

Commercial Opportunities

For partners, this model creates consulting opportunities to deliver tailored licensing guidance to help customers estimate usage and align agent deployments with business goals. It also enables partners to build and monetise their own agents, creating new revenue streams through managed services and packaged solutions.

From a technical standpoint, message consumption varies depending on the orchestration mode and the complexity of the agent. For example, generative answers typically consume two messages, tenant graph grounding consume ten, and agent actions consume five.

A Business Central Sales Order Agent interaction might consume around eighteen messages in total based on a typical conversation where:

  • A customer sends an initial request, which the agent interprets and uses to create a sales quote (1 answer + 1 agent action = 7 messages)
  • The agent then generates an outgoing message (1 answer = 2 messages)
  • Upon receipt of confirmation, the agent confirms its acceptance of the quote and generates a sales order (1 answer + 1 agent action = 7 messages)
  • The agent concludes the sequence by generating an outgoing message with an order confirmation (1 answer = 2 messages)

Partners have an opportunity to learn how to estimate usage based on customer scenarios and align licensing across the appropriate orchestration modes to inform and educate their customers. Understanding how different user licenses (e.g., Business Central, Dynamics 365, Microsoft 365) affect access is also critical.

Summary

The Microsoft monetisation model provides a flexible, scalable framework for deploying AI agents across the Microsoft ecosystem. For partners, it’s not just about understanding the model, it’s also about using it to deliver value, reduce costs, and drive innovation. With Companial’s support, partners can confidently guide customers through this new AI-powered landscape.

Partner with Companial to unlock the full potential of Microsoft’s AI monetisation model. Explore Copilot Studio Licensing, Power Platform Development and join our Special Interest Group sessions for expert insights and partner enablement support.

Chris Parkes
Latest posts by Chris Parkes (see all)

Altri post del blog