In the ERP world, "artificial intelligence" has long been a topic of discussion, often reduced to a suggestion box: it would write you a draft, provide a prediction, and you would do the rest. Microsoft Dynamics 365 Business Central's new generation of AI agents is changing this logic. Now, it's not about suggestions, it's about doing the work itself. From processing incoming orders to recording supplier invoices, AI-based agents are involved from start to finish in workflows, reducing manual burden, increasing accuracy, and enabling teams to move faster. In this article, we examine the AI agents officially introduced by Business Central one by one: what they do, how they work, where humans need to intervene, and what you, as a Turkish business, should consider before starting. As Navapp Software, we are a solution partner focused solely on Business Central; therefore, this article is written from a field perspective, not just a marketing cliché.
Let's clarify this from the start, as much of the confusion begins here. Copilot is an assistant: it stands beside you while you work on a task, suggesting text, interpreting data, and asking, "Would you like to do this?" The decision and action are yours. The AI agent , on the other hand, is designed to execute a task from start to finish on its own. It works in the background, determining the steps according to the instructions and configuration given to it, and only involves you when it needs your approval or help. In Microsoft's words, the agent conceptually interacts with Business Central just like a user; using the fields, hints, and its own instructions on the pages, it takes the next step independently. In practical terms: Copilot saves you time, while the agent takes a task entirely off your shoulders. Both work side-by-side within Business Central.
Microsoft's Business Central page highlights four tools. Three are ready-to-use "out-of-the-box" agents that handle specific finance/operations tasks end-to-end; the fourth is a platform that allows you to build your own agent.
In most SMEs, the sales process still begins in an inbox. The customer sends an email, someone reads it, finds the customer in the system, searches for products, checks stock, prepares a quote, and writes back. The Sales Order Agent is designed to automate this entire chain. The agent monitors a mailbox assigned to it. When a customer email arrives, it analyzes the subject, body, and any attachments (including PDF or image files), finds the customer among Business Central records using the sender's email address, and identifies the requested products. Then it checks the stock status; if the product is out of stock, it can even determine the earliest delivery date using a "capable-to-promise" calculation that evaluates parameters such as production capacity and lead times. If there is missing information, it conducts a multi-step email correspondence with the customer to clarify the request, finally prepares a quote with a PDF attachment, and after receiving approval, converts the quote into a sales order. There is a critical design choice here: the agent does not send any message to the customer without your approval. Every outgoing email must be reviewed by a designated Business Central user. Furthermore, the agent operates within the system with its own user ID and a limited set of permissions, sufficient only to complete its task; every action it takes is trackable in the history and records, just like the actions of other users. This is particularly meaningful for businesses involved in wholesale, distribution, and working with customers in different time zones, as the inbox is no longer limited to bu