Quick answer: AI in Microsoft Dynamics 365 Business Central now works inside everyday workflows rather than as a separate tool. Copilot features such as record summaries, drafted item descriptions and suggested bank reconciliation matches are included with Business Central licenses. Autonomous agents handle defined tasks: the Payables Agent drafts purchase invoices from vendor emails, the Sales Order Agent turns customer emails into quotes and orders, and the Expense Agent builds expense reports from receipts. A person reviews the work before anything is sent or posted. Most companies see the fastest results by starting with one high-volume workflow rather than a broad AI strategy.

Most organizations don't have an AI strategy problem inside ERP. They have a workflow clarity problem.

Teams understand that AI matters. What's less clear is where it actually changes day-to-day work without adding complexity. That gap is why many initiatives stall. The capabilities aren't missing; they just aren't tied to how work actually gets done.

What has changed over the past 12–18 months is where AI shows up. It's no longer limited to reporting or after-the-fact analysis. It's built directly into Business Central workflows, where it can surface issues, prepare next steps and reduce the effort needed to move work forward.

What AI agents are available in Business Central?

Business Central now includes two layers of AI: Copilot features that assist users inside their existing tasks, and agents that complete defined, multi-step work in the background and hand it to a person for review.

Copilot features (included with Business Central licenses)

• Summaries of customers, vendors and documents
• Drafted item descriptions and marketing text
• Suggested matches for bank reconciliation
• Chat that answers questions about your Business Central data

Business Central AI agents

• Payables Agent (generally available): Monitors a shared mailbox for vendor invoice PDFs, identifies the vendor, suggests accounts and drafts a purchase invoice. It can match invoices to purchase orders and receipts. Accounts payable reviews each draft before it moves into normal approval and posting.
• Sales Order Agent (generally available): Reads customer emails and attachments, checks item availability, creates sales quotes and converts accepted quotes into sales orders. A person reviews every outgoing customer message, and pricing or discount changes need human approval.
• Expense Agent (preview): Captures receipts, categorizes expenses and prepares expense reports. Employees and finance review reports before submission.

Organizations can also design their own agents for specific processes.

One important limit: Copilot and agents are available only in Business Central online. Companies still on Dynamics NAV or Business Central on-premises need to move to the cloud first.

How is AI changing the way work starts in Business Central?

Traditional ERP systems require users to navigate, interpret and act, and that model hasn't changed for decades. What's changing now is the starting point.

Instead of beginning with raw data or empty fields, users increasingly begin with something already in motion. A product record starts with a draft description. A reconciliation starts with suggested matches. A vendor invoice arrives as a drafted purchase document instead of a PDF someone has to key in.

The process itself hasn't been replaced. It's been compressed.

Where do Business Central AI agents save the most time?

AI delivers the most value where processes are repetitive, fragmented or dependent on manual interpretation:

• Accounts payable: The Payables Agent turns emailed vendor invoices into draft purchase invoices, so AP teams spend their time on exceptions rather than data entry.
• Sales order entry: The Sales Order Agent handles the work between a customer's email and a clean quote or order, including availability checks.
• Finance and reconciliation: Teams that reviewed transactions line by line now start with suggested matches and highlighted anomalies. The work shifts from assembling information to resolving exceptions.
• Customer service: Instead of searching across orders, invoices and emails, teams start from a summarized view of the customer.

The pattern is consistent: less time finding information and more time acting on it.

What does AI actually do before a user takes action?

Inside Business Central, AI does three things before a person acts:

1. Recognizes patterns in transactional data, such as likely reconciliation matches, outliers in financial results, or related records across customers, orders and invoices.
2. Builds context, turning raw records into a usable summary of what changed and what needs attention.
3. Prepares a next step, such as a suggested match, a drafted email or a draft purchase invoice.

None of this removes review or approval. It reduces the number of decisions needed to reach that point. Users no longer decide where to start; they decide how to proceed.

Is this full automation?

No. AI in ERP is often described as automation, but that framing can mislead.

What's really happening is a change in sequence. Instead of search, review, interpretation and then action, Business Central users increasingly begin with work the system has already interpreted and prepared. They validate, adjust and proceed. The effort before a decision shrinks, and people still make the decision.

Where AI Has the Most Impact in ERP Workflows

In Dynamics 365 Business Central, AI doesn’t deliver the most value in perfectly structured workflows. It shows up where processes are repetitive, fragmented, or dependent on manual interpretation.

In finance, that often means reconciliation and variance analysis. Teams that previously reviewed transactions line by line can now start with anomalies already highlighted and matches already suggested. The work shifts from assembling information to resolving exceptions.

In inventory and supply chain operations, the change is less about automation and more about timing. Instead of reacting to stockouts or delays after they occur, teams can identify unusual demand patterns or emerging constraints earlier, giving them more time to respond with context.

Customer-facing teams see a similar shift. Rather than navigating across systems to assemble order history, invoices, and prior interactions, they start with a summarized view of the customer. That reduces search time and makes responses faster.

Across each of these areas, the pattern is consistent: less time spent finding information and more time spent acting on it.

What AI Actually Changes Inside ERP Processes

At a functional level inside modern ERP systems, AI capabilities are doing three things before a user takes action.

First, they apply pattern recognition to transactional data. That includes identifying likely matches in reconciliation, detecting outliers in financial results, or grouping related records across customers, orders, and invoices. This reduces the need for users to manually compare and interpret individual entries.

Second, they generate structured context. Instead of presenting raw tables, the system assembles a usable summary of what changed, what stands out, and what may require attention. This is especially visible in customer activity summaries and variance explanations, where the system condenses multiple records into a single view.

Third, the system stages a next step. This might be a suggested reconciliation match, a drafted response, or a filtered set of transactions that require review. The key difference is that users are no longer determining where to start, they are deciding how to proceed.

None of these actions remove the need for review or approval. The net effect is fewer upfront decisions before reaching that point.

What do you need in place before turning on Business Central agents?

AI capabilities are no longer the limiting factor. Most organizations running Business Central online already have access. Results differ because of operational readiness.

AI depends on consistent inputs and predictable processes. When those exist, outputs are reliable. When they don't, AI exposes the gaps faster than traditional systems ever did. In our Business Central work, those gaps usually show up in reconciliation logic, posting practices, inconsistent item and customer data, and fragmented reporting.

Before turning on an agent, check three things:

• Master data: Are vendor, customer and item records clean and consistently named? Agents match what they read against what's in the system.
• Process: Is the workflow defined well enough that you could explain it to a new hire?
• Ownership: Who reviews the agent's work, and who can switch it on or off?

“Agentic” ERP Is Already Emerging - Quietly

Much of the current conversation around AI centers on the idea of agents, or AI tools that can interpret context and take action. In Business Central, elements of that model are already in place. They don’t always appear as separate tools, but they are present in how work is structured.

Users encounter them in different ways: inline suggestions that appear during tasks, prompts that generate content or answers on demand, and background processes that prepare recommendations before a user even opens a page.

They’re practical capabilities embedded into everyday workflows. And importantly, they don’t remove human control. They reduce the amount of manual effort required to reach a decision point.

Where to Start: Narrow, Not Broad

One of the most common mistakes organizations make is attempting to define a broad AI strategy before identifying a practical starting point. In reality, progress tends to begin at the process level, typically in areas where work is high-volume, repetitive, and measurable.

In practice, organizations tend to see the most progress when they approach this incrementally - an approach we’ve seen consistently across Business Central projects at JourneyTeam and elsewhere.

That often includes reconciliation, receivables follow-up, invoice review, inventory decisions, or customer response workflows. What these areas have in common is not just volume, but friction. Identify where things go wrong or slow down. AI can be applied to surface those moments earlier and provide context for resolving them. They are the parts of the business where time is routinely lost and where improvements can be measured quickly.

A More Practical Way to Think About AI in ERP

Framing AI as a transformation layer is the fastest way to misunderstand what’s actually happening. This is not a future-state upgrade. It’s a present-state divergence.

Some organizations are already operating with systems that surface issues, prepare decisions, and reduce the cost of moving work forward. Others are still relying on manual investigation as the default starting point.

That gap will not stay small. The question is no longer whether AI can be applied inside ERP. It’s whether organizations recognize—soon enough—that the cost of doing nothing is no longer neutral.

Looking for the best Microsoft Partner for a Business Central project? Check out of complete 2026 Guide on How to Choose a Microsoft Partner.

The perspectives in this article are based on real-world AI and Microsoft Copilot engagements, including work delivered by JourneyTeam across Microsoft Dynamics 365 and the broader Microsoft ecosystem.

Author Bio

Greg Crandall is the ERP AI practice director at JourneyTeam, where he works with organizations to apply AI and Microsoft Copilot in Dynamics 365 Business Central in practical, workflow-level ways—helping teams reduce manual effort, improve decision-making, and get more value from ERP.