Manual order entry and invoice processing slow down finance and sales teams every month. Microsoft has built a direct answer to this inside Dynamics 365 Business Central: AI agents that read incoming documents, match them to your records, and prepare the transaction for a human to approve.
This guide explains how AI agents in Business Central work, where they fit in a real business process, what they cost, and how to plan an implementation that doesn’t create budget surprises later.
What Are AI Agents in Business Central?
AI agents in Business Central are software components that work inside the platform using the same security and permission model as a human user. Instead of a person opening a record and typing in data, an agent monitors a mailbox or a queue, reads the incoming document, matches it against existing Business Central data, and drafts the transaction. A person still reviews and approves the output before it posts.
This is different from a Power Automate flow or a simple bot. Business Central AI agents use AI to interpret unstructured input, such as a PDF invoice or a customer email, rather than following a fixed set of rules. That is what makes Business Central agentic AI a meaningful step up from earlier automation tools, and why human review stays part of the design.
How AI Agents Work in Business Central?
Every built-in agent follows a similar pattern:
- Monitor: The agent watches a shared mailbox or a defined trigger point for new work.
- Read and interpret: It uses AI to extract data from the incoming document: line items, amounts, vendor or customer details.
- Match: It checks that data against existing records in Business Central, such as vendors, GL accounts, customers, or inventory.
- Draft: It prepares the transaction, whether that’s a purchase invoice, a sales quote, or an expense entry.
- Route for approval: A designated person, the agent supervisor, reviews the draft before it posts.
This human-in-the-loop structure runs through all of Microsoft’s current agents. None of them posts a transaction on their own. They remove the manual data entry step and leave the decision with a person.
Business Central Sales Order Agent
The Sales Order Agent watches a shared mailbox for incoming customer orders. It reads the email, checks item availability, drafts a sales quote, and converts that quote into a confirmed sales order once someone approves it.
For a business that receives a steady stream of orders by email, this removes a large share of manual re-keying and cuts the time between an order landing in the inbox, and it is reaching the warehouse.
Business Central Payables Agent
The Payables Agent targets accounts payable. It monitors a mailbox for vendor invoices sent as PDF attachments, reads the invoice content, identifies the correct vendor and GL accounts, and prepares a draft purchase invoice.
The goal is a draft accurate enough that the approver makes few or no corrections before posting. Current versions of the agent are built to handle a defined daily invoice volume, so high-volume AP teams should confirm current limits before planning a full rollout.
Business Central Expense Agent
The Expense Agent applies the same approach to expense reporting. It is aimed at employees who rarely log into Business Central directly, turning a receipt into a draft expense report without a manual form.
As of mid-2026, this agent is newer than the Sales Order and Payables agents. It is still moving through preview toward general availability, so its exact scope is worth checking against the latest Microsoft release notes before committing to it in a project plan.
Custom AI Agents in Business Central
Not every process fits a Microsoft-built agent. Agent Designer is Microsoft’s low-code tool for building custom AI agents in Business Central, letting a consultant or a power user describe a workflow in natural language rather than writing AL code from scratch. A custom agent built this way runs on the same security and permission model as the built-in Payables and Sales Order agents, which means it needs the same change management and testing discipline as any other extension to your ERP.
This matters for planning. A built-in agent’s behaviour can change on Microsoft’s release schedule. A custom agent is closer to your own code, and your team owns its upkeep, testing, and version control going forward.
AI Agents in Business Central: Use Cases
The clearest use cases today sit in two processes: order intake and payables. In practice, that covers:
- A distributor processing dozens of email-based purchase orders a day without manual entry.
- An AP team that receives vendor invoices by email and wants drafts ready for review each morning.
- A retail or services business that wants staff to submit expenses from a receipt photo instead of a form.
- A company with a repeatable but non-standard workflow, built as a custom agent through Agent Designer, that a Microsoft-built agent doesn’t cover.
Business Central AI automation is strongest where the input is document-based and repetitive, and where a human approval step is already part of the process today. It is not yet a general automation layer for every ERP task, and treating it that way is the most common planning mistake.
Business Central AI Agents Cost
Cost is where most project plans go wrong, because Business Central splits AI into two separate pricing models.
Core Copilot features, chat assistance and summarisation inside Business Central, are bundled into both the Essentials and Premium licence tiers at no extra charge.
The Payables Agent and Sales Order Agent are priced differently. They run on Copilot Credits, a consumption-based unit purchased and billed through a separate Azure subscription, metered by actual usage rather than included in the per-user licence fee. This means agent cost scales with transaction volume: a busier month for AP or sales, such as quarter-end, is also the month the agent consumes the most credits.
Three practical points follow from this:
- You need an active Azure subscription set up before either agent can go into production, not assumed as part of your existing Business Central licence.
- A custom agent built in Agent Designer runs on this same consumption-billed model once it moves past prototyping, so a power user building a new agent is also creating a new ongoing cost.
- The most reliable way to estimate cost before committing is a limited pilot against real transaction volume, rather than relying on a vendor estimate alone.
Building a single total cost of ownership figure, combining the Business Central licence and the Azure Copilot Credit estimate, before presenting a business case, avoids the most common source of budget surprise after go-live.
How to Implement AI Agents in Business Central?
Business Central AI agent implementation works best as a staged rollout rather than a single switch-on. A practical sequence looks like this:
- Confirm eligibility. Built-in agents require a Business Central cloud (SaaS) environment. On-premises and partner-hosted deployments need to move to the cloud first.
- Set up Azure and Copilot Credits. Provision the Azure subscription that will bill agent consumption before any agent goes live.
- Pick one process to pilot. Start with either the Payables Agent or the Sales Order Agent against a single entity or business unit, not a company-wide rollout.
- Run a full billing cycle. Use that pilot to get a real consumption baseline instead of a theoretical one and to see how accurate the agent’s drafts are on your own data.
- Set consumption controls. Put Azure budget alerts and spending caps in place before scaling and assign a named owner for ongoing credit monitoring.
- Extend or build custom agents. Once the pilot is validated, expand to more of the business, or use Agent Designer to cover a process the built-in agents don’t handle, with a governance review before anything reaches production.
This sequence protects two things at once: the accuracy of what the agent drafts, and the predictability of what it costs. Skipping the pilot step is the most common reason organisations end up surprised by either.
Where Business Central Agents Fit in Automation?
Business Central’s built-in agents are strong at two specific jobs: turning incoming order emails into sales orders and turning incoming invoices into payables drafts. Dynamics 365 Business Central AI agents are not yet a substitute for a broader automation strategy across your whole operation. For processes outside those two areas, a custom agent, or a combination of agents and existing Power Platform tools, is usually the right fit. Deciding which approach suits which process is exactly the kind of scoping question worth working through with a partner who implements this daily.
Get AI Agents Working in Your Business Central Environment
Reading about Business Central Sales Order Agent and Payables Agent capabilities is one thing. Getting them configured correctly, budgeted accurately, and piloted without disrupting your live finance or sales workflow is another. As a Microsoft-certified partner, we scope, configure, and support AI agent rollouts in Business Central, from your first Azure Copilot Credit estimate through to a validated production pilot and any custom agent your process needs.
If you’re evaluating AI agents for Business Central or want a clear cost estimate before you commit budget, schedule your free consultation with Business Central Partner UK now to talk through your setup and the right rollout plan for your team.
