Artificial Intelligence enters a new stage: from assistant to digital colleague
Over the past two years, artificial intelligence has become one of the most widely discussed topics in the business environment. From generating text and images to analyzing data, companies have started to discover how AI technology can save time and simplify repetitive activities.
However, most organizations face the same question:
How can artificial intelligence make a concrete contribution to a company’s operational processes?
Microsoft answers this question through a new generation of intelligent agents integrated into Dynamics 365 Business Central. Until now, Copilot has primarily functioned as an assistant that answers questions and provides suggestions. The new generation of AI Agents goes much further. These agents do not simply provide information. They execute complete processes.
The first of these is Sales Order Agent, an intelligent agent that automates the process of receiving and processing orders received by email.
What is Sales Order Agent?
Sales Order Agent is an AI agent integrated into Microsoft Dynamics 365 Business Central, designed to automate one of the most time-consuming activities in sales departments: processing customer orders. Unlike the traditional process, in which an operator reads each email, identifies the customer, searches for the requested products, checks their availability, and manually creates a quote or sales order, all these steps can be handled by the AI agent. In practice, Sales Order Agent acts like a digital colleague that continuously monitors a dedicated email inbox and starts working as soon as it receives a customer request.
How does it work in practice?
Let’s assume that one of your customers sends the following message: “Hello, I would like a quote for 50 units of product X and 20 units of product Y, with delivery next week.”
In a traditional process, this request involves several steps:
- someone reads the email;
- identifies the customer;
- searches for the products in the ERP;
- checks stock availability;
- calculates prices;
- prepares the quotation;
- sends the response to the customer.
Sales Order Agent automatically takes over these activities and analyzes the content of the message using artificial intelligence, identifies the relevant information, and starts processing the request.
First, it understands the customer’s request
The first thing Sales Order Agent does is interpret the incoming message. It does not simply look for specific keywords. It uses AI models to understand the customer’s intent.
The agent automatically identifies:
- who the customer is;
- which products they are requesting;
- the requested quantities;
- the delivery date;
- any external references;
- other relevant information included in the message.
If some details are missing or ambiguous, the agent does not simply stop.
It asks questions just like a sales team colleague
One of the most impressive capabilities of Sales Order Agent is its ability to communicate through email. If the customer’s request does not contain all the necessary information, the agent can automatically generate a response asking for clarification. For example:
- which product is being requested;
- which product variant the customer wants;
- which unit of measure should be used;
- what the delivery address is;
- when the customer wants the delivery.
These conversations can continue until all the information required to prepare the commercial document is available.
It automatically checks product availability
Once the products have been identified, the agent consults the information available in Business Central. It can check:
- whether the products exist;
- stock availability;
- available quantities;
- the commercial information required to process the request.
If certain products are unavailable, the customer can be informed before the commercial document is issued. This reduces the number of email exchanges and accelerates the entire sales process.
It automatically generates the quote or sales order
Depending on the configuration selected by the company, Sales Order Agent can create:
- a sales quotation;
- or directly a sales order.
The document is automatically generated in Business Central and can be sent to the customer as a PDF, including the requested products, quantities, prices, taxes, units of measure, delivery terms, and other relevant information.
AI does not replace people. It helps them work more efficiently.
An important aspect is that Microsoft is not aiming to eliminate human control. Sales Order Agent can operate autonomously for many activities, but organizations can configure the points at which users need to intervene. For example, the sales team can review messages generated by the agent before they are sent to the customer or approve commercial documents before they are issued.
In this way, artificial intelligence takes over repetitive activities, while people remain focused on customer relationships and commercial decisions. But what exactly happens behind the scenes?
One of the interesting aspects of Sales Order Agent is that it does not follow a rigid, predefined path for every request. The agent uses AI capabilities to interpret the context and interacts with Business Central in a way that is similar to a human user.
In other words, this is not simply an automated workflow based on the logic of “if A happens, execute B.” The agent analyzes the situation and determines the steps required to achieve its objective.
The Agent navigates Business Central like a user
Sales Order Agent starts from the Role Center configured for it and uses the information available in the Business Central interface to determine which actions need to be performed. The agent can navigate between pages, take actions, and enter data using the metadata available in the application, such as names, tooltips, and other interface properties. This approach gives the agent a certain degree of flexibility. For example, if an organization uses custom fields or actions that are relevant to the sales process, the agent can also interact with them, depending on the context. It can also attempt to resolve certain validation errors automatically by analyzing the messages displayed by Business Central and adjusting the information entered.
This is an important difference compared with traditional automation, which generally operates based on predefined steps and rules.
Where does it get its information?
For Sales Order Agent to process a request correctly, it needs access to the relevant information in Business Central. The first step is identifying the customer. The agent uses the sender’s email address to search for the associated contact in Business Central. It then checks quotations and orders associated with that customer to ensure that documents belonging to different customers are not mixed up. If the sender is identified as an existing customer, the agent works with the documents associated with that customer. If there is a contact but that contact is not registered as a customer, the agent can, in certain scenarios, work with documents associated with the contact. If the email address cannot be identified, user intervention may be required to update the information or register a new contact. This mechanism is important for control and traceability: the agent needs to know who the request is being processed for before modifying or creating documents.
Finding products: data quality is extremely important
Another essential step is identifying the products requested by the customer. The customer will not always use the exact name stored in the ERP. They may provide a partial description, a vendor item number, a product characteristic, or a commercial name used internally by the company. Sales Order Agent analyzes the content of the email and searches for the products using the information available in Business Central. The fields it can use include:
- item number;
- description;
- extended description;
- search description;
- GTIN – Global Trade Item Number;
- vendor item number;
- product variants;
- item references;
- item attributes;
- item category;
- translations;
- product identifiers.
This is where a very important aspect emerges for any company looking to use AI in its ERP: Data quality becomes essential.
If product information is incomplete, unclear, or inconsistent, the agent has less information on which to identify the right product. That is why implementing an AI agent does not necessarily start with AI. It can start with a much simpler question:“How well organized is our data?”
Product descriptions, attributes, categories, and additional information can all contribute to the agent’s ability to identify the requested products.
Checking product availability
After identifying the products, Sales Order Agent checks their availability. The analysis is not limited to the quantity currently available in stock. The agent can take multiple parameters into account, including the requested quantity, desired delivery date, location, as well as planned or scheduled incoming quantities. This allows the company to provide a realistic response to the customer more quickly. Instead of an email exchange such as:
“We have the product.”
“But when can you deliver it?”
“I need to check.”
“And what about the other products?”
“I’ll get back to you.”
the agent can centralize the relevant information as part of the quotation preparation process.
From request to PDF quotation
Once the customer, products, and availability have been identified, Sales Order Agent can create the sales quotation. The quotation is generated in Business Central and prepared for delivery to the customer as a PDF document. It can include the requested products, quantities, units of measure, prices, taxes, requested delivery date, and the external reference provided by the customer, along with other relevant information.
For sales teams, the benefit is clear: a significant part of the administrative work required to prepare a quotation is handled by the agent.
The conversation does not stop after the quote is sent
One of the things that differentiates Sales Order Agent from traditional automation is that the process can continue even after the quotation has been sent. The customer can reply by email and request changes, modify the quantity, remove a product, request a different delivery date, ask for another configuration, and much more. Sales Order Agent can identify the existing quotation and help update it based on the customer’s requests.
In practice, the sales process can continue conversationally through email, without a team member having to manually restart the entire process.
And what happens when the customer says “Yes”?
Once the quotation has been approved and the order has been placed, Sales Order Agent can convert the quotation into a sales order.
It can then prepare the confirmation message to be sent to the customer.
This is where another important element of the Sales Order Agent architecture comes into play:
The human remains in the loop.
Designated users in Business Central must review and approve messages sent to customers. They can also review the actions performed by the agent and the documents it created. This model combines automation with human control. AI handles information processing and repetitive activities, while the user retains control over external communication and important decisions.
What needs to be prepared before activating the Agent?
For Sales Order Agent to function correctly, the organization needs to prepare several elements. First, an email inbox needs to be configured for the agent to monitor. The administrator then defines preferences regarding how the agent operates and which users are authorized to use it. The agent is assigned permissions and a Business Central profile, and the administrator can adjust these rights according to the organization’s needs. This aspect is essential. An AI agent should not have unlimited access to data and processes. Just like any other user or service in an ERP system, its access should be configured according to its role and responsibilities.
AI is important, but it needs a well-organized ERP system
Sales Order Agent highlights one of the important transformations AI is bringing to ERP systems. But it also demonstrates something equally important:
AI cannot compensate indefinitely for poor data or unclear processes.
The better structured the information in Business Central, the better context the agent has to perform its tasks.
Therefore, a discussion about Sales Order Agent should also include data quality, sales processes, permissions, Business Central configuration, and the way the team currently works.
How much does it cost to use Sales Order Agent?
Another aspect that needs to be considered is the consumption model.
Sales Order Agent uses Microsoft Copilot Studio messages for AI interactions, which involve costs depending on the complexity of those interactions. To use the agent, a billing model needs to be configured for the Business Central environment. Therefore, when evaluating the implementation of Sales Order Agent, the estimated volume of interactions should also be taken into account.
The right question is not simply “How much does the agent cost?” It is, above all: “How much does manually processing customer orders cost us today?” Employee time, email exchanges, product checks, quotation preparation, and the risk of errors all have a cost. And this is where the real potential of Sales Order Agent lies.
Now, we will move beyond the technical perspective and look at Sales Order Agent from a business perspective: what the sales team gains, which types of companies can benefit from it, how human control is maintained, and what should be analyzed before implementation.
We have already looked at how Sales Order Agent works, from the moment a customer sends a request by email to identifying products, checking availability, generating a quotation, and converting it into an order. It’s time we take a step back and look at the technology from the company’s perspective.
Because besides “What can Sales Order Agent do?”, the more important question is: “What changes in my team’s day-to-day work if I use such an agent?”
The answer is connected to a word we hear more and more when discussing AI: autonomy.
From automating a task to automating a process
Traditional automation generally works based on clearly defined rules. If A happens, the system does B. If a document meets certain conditions, a specific action is executed. Sales Order Agent introduces a different approach. The agent receives an objective and uses the available information to determine the steps required to achieve it. Microsoft explains that the agent uses natural-language instructions, the data displayed in Business Central, and interface metadata to determine which actions need to be performed. This distinction is important. We are no longer talking about automating a single activity, such as entering an order. We are talking about automating an entire workflow that can include interpreting the email, identifying the customer, searching for products, checking availability, creating the quotation, communicating with the customer, and eventually converting the quotation into an order.
What does the sales team gain?
For a sales department, one of the biggest challenges is balancing administrative activities with those that directly contribute to customer relationships. A sales representative can spend a significant part of the day:
- reading and classifying emails;
- searching for products in the ERP;
- checking prices and availability;
- manually entering information;
- preparing quotations;
- updating documents;
- tracking changes requested by customers.
Sales Order Agent can take over a large part of this workflow. The team remains involved where decisions, verification, or intervention are required, while the agent can work in the background on repetitive activities.
This can change how time is allocated within the team: less time spent on administrative operations and more time dedicated to customers.
An important advantage: response speed
In a sales process, response time matters. A customer who sends a request naturally expects a response. In the traditional scenario, response time depends on the availability of the colleague handling the request, their workload at that particular moment, and the number of checks required. Sales Order Agent can continuously monitor the configured email inbox and start processing identified requests. This means that processing a request is no longer as dependent on whether a particular colleague is available at that exact moment.
For companies receiving a high volume of sales inquiries, this difference can be significant.
But AI does not mean “let the Agent do anything”
This is one of the most important aspects of Sales Order Agent. Although the agent is designed to operate autonomously in the background, Microsoft keeps the human user involved in the process. Designated users can review the steps performed by the agent and the documents it creates. Moreover, messages sent to customers must be reviewed and approved by designated users before they are sent. This creates an interesting collaboration model: The AI Agent executes. The human reviews and decides.
This approach may make automation easier for organizations to adopt because automation does not mean giving up control.
Control remains where it matters
Let’s assume the agent correctly identifies the products and creates the quotation. A user can review the document and intervene, for example, to add a discount or update shipping costs. Microsoft mentions the possibility for users to provide the agent with additional information when required. This mechanism is particularly important in companies where prices, discounts, or commercial terms require human validation.
AI can accelerate the process, but the commercial decision remains with the company.
Which companies can benefit the most?
Sales Order Agent is particularly useful for organizations that frequently receive orders or quotation requests by email and have a relatively well-structured sales process. For example, it may be relevant for companies that:
- handle a high volume of requests;
- sell standardized products;
- have well-structured commercial information in Business Central;
- frequently receive orders by email;
- spend significant time manually entering orders;
- need faster response times;
- want to reduce the administrative workload of the sales team.
The higher the volume of requests, the more interesting the potential for efficiency gains becomes.
Good Data makes a better AI Agent
One of the most important lessons from Sales Order Agent is that artificial intelligence needs good context. Microsoft explains that the agent can identify products even when the descriptions received from customers are incomplete or vague. However, search efficiency is influenced by the quality of the information available in Business Central. Product descriptions, attributes, categories, translations, item numbers, and additional information can help the agent find the right product. That is why, before implementing an AI agent, it is worth assessing the quality of the data already available in the ERP.
A high-performing agent working with incomplete data will still be limited by that data.
Sales Order Agent is not a separate ERP Project
Another important aspect is that the agent does not operate in isolation. It works directly with the information and functionality available in Business Central. Customers, products, quotations, orders, inventory availability, and commercial information are all part of the ERP system. This means that the value of the agent is closely connected to how Business Central is configured and used.
If sales processes are clear and data is well structured, the agent has a better context in which to operate.
Security and permissions remain essential
Any discussion about AI in an ERP system must also include security. Sales Order Agent operates based on the permissions and profile assigned by the administrator. These permissions determine the areas of Business Central the agent can access, and administrators can modify them according to the organization’s needs. Therefore, implementation needs to be approached with the same level of attention given to any other user or process interacting with company data.
AI can automate processes, but it should not bypass the organization’s access and control rules.
A new way of looking at productivity
Perhaps the most important change brought by Sales Order Agent is a change in perspective. Instead of asking: “How many orders can an employee enter in a day?” we can start asking: “Which activities in the order-processing workflow still need to be performed manually?” It is a subtle but important difference.
If the AI agent can take over repetitive activities, people can focus on tasks that require experience, creativity, negotiation, and direct customer interaction.
How much does it cost? And, more importantly, how much does it cost not to automate?
Sales Order Agent uses Microsoft Copilot Studio messages for AI interactions, which involve a cost depending on the complexity of those interactions. Using the agent requires a billing model to be configured for the Business Central environment. Therefore, any evaluation should start with the actual volume of activity. For example:
How many requests do you receive each month?
How many minutes or hours are required to process each one?
How many people are involved?
How much time is spent on checks and email exchanges?
What is the cost of an error?
And perhaps most importantly: How much is the time worth that the team could otherwise spend on sales activities?
These are some of the questions that can determine whether automation makes sense for a particular organization.
What are the first steps in implementing Sales Order Agent?
The implementation of Sales Order Agent should not be viewed as a race to introduce AI at any cost. The first step is to understand the current process. Then, several areas can be analyzed:
- Request volume – How many orders and quotation requests are processed by email?
- Order complexity – How standardized are the products and commercial terms?
- Data quality – Are product and customer information well structured?
- Existing processes – Where do most manual activities currently occur?
- Permissions and control – What actions can the agent perform?
- Expected benefits – How much time and how many resources can be saved?
After this analysis, it can be determined whether Sales Order Agent is suitable for the specific process and how it should be configured.
Artificial Intelligence working together with people
Sales Order Agent shows us an important direction in the evolution of ERP systems. This does not necessarily mean that people will disappear from processes. It means that their role can change. The agent can monitor, interpret, search, verify, prepare documents, and conduct conversations. People can review, decide, negotiate, and build customer relationships, while Business Central becomes the place where these activities come together.
In conclusion, Sales Order Agent transforms a process that was previously largely administrative into a workflow in which Artificial Intelligence can take over a significant part of the operational work. From the first email to the final order, the agent can contribute to identifying the customer, finding products, checking availability, preparing quotations, managing changes, and sending documents to the customer. At the same time, users retain control over the steps that require verification and approval.
For companies, the potential benefit goes beyond simply reducing the number of manual operations. It is the opportunity to build a faster, more scalable sales process that is better connected to the ERP.
Sales Order Agent is a concrete example of how AI is beginning to move from the role of a conversational assistant to that of an agent capable of executing business processes.
For organizations already using Microsoft Dynamics 365 Business Central, the question is becoming increasingly interesting: What other repetitive processes could be taken over by AI?
At Arggo, this is one of the directions we are exploring as we analyze the evolution of Business Central and how new AI capabilities can be transformed into concrete business benefits. If you would like to evaluate how well Sales Order Agent fits your sales processes, the Arggo team can help you identify the scenarios where automation can deliver the greatest value.
