Which AI agents are there in Business Central?

Business Central comes with several agents: a Sales Order Agent for order capture, a Payables Agent for vendor processing, finance-related agents around reconciliation and posting, an agent for customer communication, plus the Agent Designer for your own agents. Sensible first is the agent for the process that is most cleanly regulated with you.

Which agents exist today?

Each agent covers a delimitable process. What matters is not what the agent can do in theory, but how data-mature the process already is with you.

• Sales Order Agent: takes in incoming orders, assigns items and terms and prepares the order. Strong when the item master and the pricing logic are clean, error-prone when they are not.
• Payables Agent: processes incoming invoices, matches them against the purchase order and the goods receipt and prepares the posting. Needs a clean procurement setup in order to genuinely relieve the workload.
• Finance-related agents: support reconciliation, assignment and recurring posting routines. They live on unambiguous account rules.
• Customer Communications: supports recurring customer communication, for instance order confirmations or status updates, along your own specifications.
• Agent Designer: the no-code layer with which you configure your own agents for your own flows, instead of being limited to the standard agents.
• Expense Agent: accepts expense receipts by photo, web app or email, extracts them, assigns categories and hands reports over for approval. Available in Germany and Austria since summer 2026, 50 credits per receipt.

The youngest of the three is the Expense Agent, available in German since summer 2026. What it takes over, what a receipt costs and what it needs from your master data is in What does the Expense Agent do in Business Central?

Source: Microsoft Learn: Sales Order Agent overview

Where do you start?

The best first agent is not the one that promises the most but the one whose process is most clearly governed and data-ready at your company. That is where the benefit is measurable and the risk manageably small. An agent on an unclean process mainly speeds up the errors.

A workable selection method takes an hour: take the five agents, rate for each process the volume, the clarity of the rules and the master-data quality, and start where all three are good. If in the end none is left, that is the real answer and the pointer to where to work first.

AI without hype means: the agent works on your data foundation, not in its place. It presupposes a governed process, it does not replace one. If you want to work out which agent will genuinely carry first at your company, we go through your processes together in a conversation.