Commercial

Business Central AI billing in Australia: Copilot included vs agent consumption cost

A practical Australian buyer guide to where Business Central Copilot is included, where autonomous agents create extra consumption cost, and how to budget pilots without getting surprised later.

The confusion is understandable because Microsoft presents these ideas in two different ways. The Business Central pricing page says Copilot is included, yet Microsoft's current Business Central and Copilot Studio documentation also says that agent scenarios such as Sales Order Agent, Payables Agent, and Expense Agent use consumption billing through Copilot Credits or pay-as-you-go setup.

For Australian SMB buyers and existing customers, the practical job is not to argue about whether AI is strategic. It is to separate included user-assistance features from billable autonomous or background agent work, then budget the first pilot with real transaction volume and real exception handling in mind.

What Microsoft officially separates

• Microsoft's Business Central pricing page says customers get Business Central with Microsoft Copilot included. That is the starting point, but it is not the full commercial answer.
• The same pricing page also points to a Copilot Studio plan, a pay-as-you-go option, and separate Dynamics 365 agent licensing paths. That means “Copilot included” should not be read as “every AI workflow runs for no extra cost”.
• Microsoft's Business Central administration documentation says some capabilities are billed by consumption and explains how to link a Business Central environment to a Power Platform environment to enable that billing.
• Microsoft's Business Central requirements documentation then assigns Copilot Credit consumption rates to specific agent activities. That is the detail buyers need when moving from curiosity to pilot design.

Which current Business Central agent scenarios are consumption-billed

• Sales Order Agent is billed through Copilot Credits. Microsoft's official example for 100 monthly interactions totals 1,650 credits based on the mix of customer-intent analysis, customer lookup, item or stock checks, quote creation, and order conversion actions.
• Payables Agent is also billed through Copilot Credits. Microsoft's official example for 100 invoices with three lines each totals 6,500 credits because the workflow combines email understanding, draft invoice creation, line extraction, and line-item processing.
• Expense Agent preview is listed in the same Business Central billing documentation at 50 Copilot Credits per expense receipt processed.
• The practical lesson is that cost is driven by work performed, not only by named user count. Teams should budget around documents, order threads, and exception-heavy scenarios instead of assuming AI cost scales neatly with headcount.

Why Sales Order Agent and Payables Agent cost differently

• Microsoft's own examples show that AP-style document handling can consume meaningfully more credits per work item than order-intake assistance because invoice extraction is line-intensive and document-led.
• Sales Order Agent economics depend heavily on the quality of customer matching, item master data, availability checks, and whether many email threads end in quote updates before order conversion.
• Payables Agent economics depend on invoice volume, average line count, attachment quality, and how often supplier or PO exceptions force rework.
• This is why two businesses with the same Business Central user count can face very different AI run-rates. The driver is process shape, not only licence inventory.

What the Microsoft 365 Copilot fair-use note does and does not mean

• Microsoft's Copilot Studio licensing guide says there is no separate Copilot Studio message charge when a licensed Microsoft 365 Copilot user accesses an agent in a Microsoft 365 app and the agent operates only on behalf of that authenticated user within Microsoft 365 services, subject to fair-use terms.
• That note is relevant for employee-facing chat and assistance patterns. It is not a safe assumption for Business Central agents that monitor shared mailboxes, run background tasks, or execute ERP-specific actions with dedicated billing guidance in Business Central documentation.
• Inference from Microsoft's own documents: if Business Central publishes a consumption-billing setup path and per-scenario credit examples for an agent capability, buyers should budget that capability explicitly even if some other Copilot experiences are bundled elsewhere in Microsoft 365.

How Microsoft lets teams pay for it

• Microsoft documents two main ways to fund usage: prepaid Copilot Credit packs and pay-as-you-go billing tied to an Azure subscription through Microsoft Power Platform.
• Business Central administrators must link the Business Central environment to a Power Platform environment before consumption billing can run. This is an implementation prerequisite, not a later admin tidy-up.
• Because credits are a shared capacity model, the finance question is not just “what is the list price of a pack?” It is “which environments, agents, and business processes are allowed to consume that shared pool?”.
• A sensible rollout therefore needs ownership for budget guardrails, environment governance, and monthly usage review before teams enable more than one agent scenario.

A practical budgeting method for Australian buyers

• Step 1: separate AI assistant value from AI agent value. Included Copilot features can still matter, but the budget risk usually starts once mailbox-driven or background agent workflows enter production.
• Step 2: estimate monthly work items instead of users. Count invoices, order-request threads, receipts, and the likely exception rate for each process you want to pilot.
• Step 3: model best-case and messy-case usage. Microsoft's published examples are useful baselines, but real Australian operations usually include partial deliveries, supplier exceptions, customer-specific pricing, and attachment noise.
• Step 4: pilot one process first and measure credits per successful outcome. For most teams that means either AP draft creation or email-driven order capture, not both at once.
• Step 5: decide who owns the commercial review. AI usage inside ERP should sit with finance and process owners as well as the Business Central admin, because ongoing cost follows operating behaviour.

Questions buyers should ask their partner or Microsoft now

• Which current Business Central AI features are fully included with our existing licences, and which ones trigger Copilot Credit or pay-as-you-go consumption?
• What is the cleanest first pilot in our environment based on real process volume and master-data quality: Sales Order Agent, Payables Agent, or something else?
• How will usage be monitored by environment, by agent, and by month so finance can distinguish experimentation from steady-state operating cost?
• What fallback process exists if we pause usage because cost, exception rate, or control concerns are higher than expected?

What Australian buyers should conclude now

• Business Central with Copilot included is real, but it does not mean every AI-driven workflow is bundled into the base ERP subscription.
• The current Microsoft documentation is clear enough to justify a dedicated AI budget line for agent pilots, especially around Sales Order Agent, Payables Agent, and future document-heavy scenarios.
• The safest commercial approach in Australia is to treat AI consumption like any other ERP operating cost: define the process scope, measure usage, review exceptions, and only then scale it across entities or teams.

FAQ

• Is Copilot included in Business Central? Microsoft's pricing page says yes, but that does not remove separate consumption billing for all agent scenarios.
• Are Sales Order Agent and Payables Agent free once we own Business Central licences? No. Microsoft's current Business Central documentation gives explicit Copilot Credit billing examples for both.
• Should buyers budget AI by user count? Not primarily. Microsoft's billing examples make it clear that documents, order threads, and processed actions are the better first budgeting unit.
• Does this matter only for large enterprises? No. SMB teams can feel AI usage-cost surprises faster because they usually pilot with tighter budgets and less overhead for monitoring.