What are AI agents in Microsoft Business Central?

AI agents in Business Central are autonomous functions that complete a task on their own inside the system, rather than only making suggestions the way Copilot does. They read your data, decide along rules and carry out steps, for example in Sales Order, Payables or Finance. The difference from the neighbouring terms matters in practice: Copilot suggests, a person decides and approves. An agent acts on its own within given rules and presents the result for approval. Classic automation follows rigid rules without context and stops as soon as a case does not fit the pattern. An agent is therefore not a chatbot; it works inside the system, on your data and by your rules. Wave 1 2026 brings native agents and the Agent Designer for building your own. The benefit comes not from the technology but from deciding which tasks an agent may take on. Drawing that line is governance work.

Where this leads in the medium term, in three theses: ERP transformation 2027

Agent, Copilot or automation: what is the difference?

Three terms get mixed up in the market, and the difference matters in practice. Copilot proposes, a person decides and approves. An AI agent carries out a task independently within given rules and submits the result for approval. Classic automation follows rigid rules without understanding context and stops as soon as a case does not fit the pattern.

An agent is therefore not a chatbot. It works inside the system, on your data and by your rules. Microsoft documents the available AI capabilities in Business Central and likewise separates assistance from autonomous agents.

Source: Microsoft Learn: Configure Copilot and agent capabilities

In practice this means the value does not come from the technology but from deciding which tasks an agent may take on and where the limit lies. Defining that limit is governance work, not IT work, and it belongs before the configuration.

What does an agent need to function?

An agent needs three things, and none of them is technology. First clean master data, because it does not distinguish between a well-kept record and one that has grown over the years. Second clearly defined processes, because it can only execute what has been decided. Third defined approval and escalation paths, so it is clear who steps in when a case falls outside the rule.

If one of these three is missing, the agent scales the error rather than the relief. A single manual mistake becomes an automated serial mistake, and it surfaces later because nobody reviews every transaction any more. That is exactly why the order is data foundation, process, agent, and not the other way round.

The Sales Order Agent illustrates the pattern well: it captures orders from incoming enquiries and submits them for approval. It replaces the data entry, not the responsibility. The bottleneck moves from input to review and data upkeep.

Source: Microsoft Learn: Sales Order Agent overview