Dynamics 365 Business Central agent capabilities: how much does they cost me?

With Dynamics 365 Business Central 2025 Wave 1 release (version 26) Microsoft officially launches the Sales Order Agent, the first agentic AI functionality built-in into the platform that automates processing sales quotes and orders starting from customer email requests.

The agent uses AI to analyze customer requests received via email, locate the customer in Business Central, and engage in multi-turn email conversations to clarify requests if important details are missing or more choices are available. It also checks and informs the customer about the availability of the items they’re looking for and follows up with a sales quote. The quote is formatted as a PDF and includes the requested items, quantities, units of measure, prices, taxes, requested delivery dates, provided external document reference, and other important details.

Here is a Microsoft diagram explaining the process in details:

This is only the first of many other agents that Microsoft is planning to introduce into the core platform.

The Sales Order agent uses Microsoft Copilot Studio messages for AI interactions and this means that you will be charged accordinlgy to the interactions with the AI agent you have during time.

To activate billing, you need to link your Business Central environment to a Power Platform environment.  Linking to a Power Platform environment lets the Business Central environment inherit selected settings from the Power Platform environment. It also provides a default target environment when setting up features that integrate Business Central with Power Platform and other Dynamics 365 products.

Please remember that links between Business Central environments and Power Platform environments are exclusive: you cannot link multiple Business Central environments to a single Power Platform environment or vice versa.

To link your Business Central environment to a Power Platform environment, open the Business Central tenant’s Admin Center, select your environment and in the Linked Power Platform Environment section on the environment page, select Link under Environment:

Now select your Power Platform environment and confirm the link. Please note that the Business Central environment will be restarted after the confirmation:

Only Power Platforms environments meeting the following conditions are available to link to your Business Central environment:

• In the same Azure Geo.
• Same type (Production or Sandbox).
• Not already linked to another environment.

When you have the environment linked, you can setup your agent billing model.

Business Central supports two billing models: prepaid capacity and pay-as-you-go. The prepaid capacity model uses Copilot Studio message pack subscriptions, which are a licensing option for Microsoft Copilot Studio that you purchase in advance. The pay-as-you-go model charges for the actual number of messages consumed by agents during the month. 

You can have both billing models active on your environment and in that case the prepaid capacity is consumed first.

Prepaid capacity setup

To activate a prepaid capacity billing, sign-in to Power Platform Admin Center (here I’m using the new UI) and on the navigation pane, select Licensing. In the Licensing pane, under Products, select Copilot Studio. From here you can acquire and manage your prepaid capacity plan:

Pay-as-you-go setup

If you want to go for the pay-as-you-go plan, create a billing plan and simply link the billing plan to your Azure subscription:

Ok, and now what an agent can cost me?

There are three different Copilot Studio license plans actually available:

• Pay-as-you-go: $0.01 per message.
• Message packs: 25000 messages per month, $200 per tenant, per month.
• Copilot Studio in Microsoft 365 Copilot: $30 per user, per month.

With the pay-as-you-go plan, at the end of each month, an organization only pays for the actual number of messages its agents consumed during the month (no up-front license commitment is required).

With the message pack plan, one message pack equals 25,000 messages a month.

Messages are a measure of the time and effort required for your agent to retrieve information and respond to prompts and any actions or custom skills the agent uses. The number of messages taken into account for each response or action depends on the complexity of the task completed by the agent. Here is the official actual Microsoft’s table regarding messages consumption and its related billing:

When checking this table, remember the following:

• Autonomous triggers refer to events or conditions that automatically initiate an agent to take action, without requiring a user to manually invoke it.
• Classic answers are predefined responses manually authored by agent makers. They’re static and don’t change.
• Generative answers are dynamically generated using AI models. They can adapt and change based on the context and the knowledge sources they’re connected to.

Calculating the number of consumed messages by an agent process is not an easy task, but I want to do a quick calculation starting from an Excel file that the Power Platform community shared some weeks ago (thanks http://x.com/ChrisO_Brien). I’ve modified it a bit in order to consider also autonomous actions.

Let’s play a bit with some scenarios…

Scenario A: 1 autonomous agent using 1 autonomous action (trigger) and 5 average AI-generated answer per conversation, used by 10 users per day: total cost is around 77$ per month and it can consume around 350 messages per day.

Scenario B: 1 autonomous agent using 1 autonomous action (trigger) and 8 average AI-generated answer per conversation, used by 15 users per day: total cost is around 108$ per month and it can consume around 490 messages per day:

Scenario C (let’s grow up ): 1 autonomous agent using 2 autonomous actions and 10 average AI-generated answer per conversation, used by 20 users per day: total cost is around 198$ per month and it can consume around 900 messages per day:

Scenario D (let’s grow up again): 3 autonomous agents using 2 autonomous actions and 15 average AI-generated answer per conversation, used by 30 users per day: total cost is around 308$ per month and it can consume around 1400 messages per day:

Scenario E (let’s go enteprise): 3 autonomous agents using 5 autonomous actions and 20 average AI-generated answer per conversation, used by 60 users per day: total cost is around 803$ per month and it can consume around 3650 messages per day:

These are just simulations and can not reflect a real cost (it depends a lot on how the agent is created and on what the agent is doing) but I think that they can give you an idea.

Conclusion

To determine what a Business Central agent can cost, it’s important to determine the number of messages it will consume for executing a starnadrd process.

For example, if your agent processes 100 emails per month and each email interaction involves:

• 3 generative answer (2 messages each)
• 1 agent action (5 messages)
• 1 tenant graph grounding (10 messages)
• 10 agent flow actions (13 messages per 100 actions)

The total messages per email interaction would be:

• Generative answer: 6 messages
• Agent action: 5 messages
• Tenant graph grounding: 10 messages
• Agent flow actions: 1.3 messages (for 10 actions)

Total messages per email interaction: 6+5+10+1.3=22.3 messages.

For 100 emails per month: 22.3×100 = 2230 messages.

Considering the Message packs plan at $200 per tenant (with 2500 messages included) a process like this could consume all your pre-paied pack.

Is it cheap? Not always… but when talking about agentic AI and related costs, I think you always need to remember an aspect: the return of investment (ROI).

For many repetitive tasks, an AI agent can replace a human. It can help on automating routine tasks and this automation frees up valuable time for your employees to focus on more strategic activities. The agent can also engage in multi-turn email conversations with customers to clarify requests and ensure accurate order processing (also avoiding time consuming activities for a call center). By automating tasks that would otherwise require manual intervention, you can reduce the need for additional staff. The agent can provide quick and accurate responses to customer inquiries, improving customer satisfaction and loyalty.

Are all these aspects a beneficts for your business? Are the associated agentic AI costs sustainable for you accordinglu to the previously mentioned advantages? If so, then you’re a good candidate for using AI agents in Business Central. If no, you can always turn them off .

Before closing this post, I want to remember an important aspect: here we’re talking about out-of-the-box AI agents provided by Microsoft in the product or AI agents created by using low-code tools like Copilot Studio. Remember that for complex AI projects that require high levels of customization and scalability, a full-code approach for your agentic AI solution may be more suitable and it can lead to a final architecture with a much lower cost.

Obviously, for that we need “pro AI developers”, but as I always love to emphasize:

the world of pro code still exists and if you want a scalable, limitless and maybe even AI model-independent solution, writing good, healthy code with the languages ​​that support the main AI frameworks is a good and right thing to do.

If you will be at BC TechDays in less than 2 months, we’ll see some of these “pro code” tools for creating agentic AI solutions. If interested on this topic, just bookmark this session (and start saving money) :

Original Post https://demiliani.com/2025/04/16/dynamics-365-business-central-agent-capabilities-how-much-does-they-cost-me/