What AI in Dynamics 365 Business Central Does for Manufacturing Planning
What ships today, what's in preview, and what you build for forecasting, planning and order promising.
If you run Dynamics 365 Business Central in a plant, you've heard "Copilot" in every roadmap session this year. The more useful question is narrower: which of these features help your planner answer "can we ship it by the 15th?" Here's our read as of late September 2026, sorted by status.
Ships today: inventory forecasting on Azure AI
The Sales and Inventory Forecast extension reads posted sales from item ledger entries and uses a preconfigured Azure AI service to predict demand by item. It flags items likely to stock out and suggests quantities when you create purchase orders. Microsoft recommends monthly periods over a 12-month horizon.
Know the limits before you promise it to your team. Forecasts are aggregated at item level, not by location or variant. Periods with big swings hurt accuracy. And it predicts sales demand; production still plans through MRP.
Ships today: Sales Order Agent, item insights and substitutes
The Sales Order Agent reads customer requests from a mailbox and drafts quotes and orders for a person to review. Since July 2026 you can run more than one per company. Copilot item insights add a plain-language summary and an Item Statistics page with inventory value, sales growth and margin by period. Item substitute suggestions propose alternates when a component is short.
In preview: asking your data questions in plain language
Chat with Copilot and analysis assist let a planner ask for open production orders due this week, or group released orders by item, without building a report. Both are production-ready previews under Microsoft's supplemental terms. A unified Copilot chat experience is also in preview with update 29.0.
Still rules-based, and that's a good thing
The planning worksheet runs MPS and MRP against your BOMs, routings, lead times and stock. Order promising gives you available to promise from existing supply and capable to promise for the shortfall. None of this is AI. It's predictable and auditable, which is exactly what you want under a customer commitment.
Update 29.0, in preview now and due for general availability in October, adds releasing production orders straight from the planning worksheet and easier work with capacity calendars.
Where native planning stops: basic finite loading covers capacity-constrained resources. Detailed sequencing, changeover optimization and shop floor data collection at scale usually call for an ISV extension. We'd rather tell you that up front.
You build: agents on the Dynamics 365 Business Central MCP server
The MCP server has been generally available since April 2026. It lets agents you build in Copilot Studio read and act on your ERP data. Examples we'd scope with a manufacturer: a morning shortage brief that checks tomorrow's production orders for missing components, or a watcher that flags purchase orders whose new receipt date puts a promised ship date at risk. These are projects, not features you switch on.
Readiness decides how much of this pays off
Nearly 98% of manufacturers are exploring AI and only 20% feel fully prepared. The difference is almost always data: posted sales history, trusted lead times, certified BOMs and routings, and output recorded when it happens.
It's also a staffing question. The Deloitte and Manufacturing Institute 2025 survey found 46% of manufacturers struggle to fill planning and scheduling roles. Getting the first draft of a forecast from the system frees scarce planners for the calls only they can make.
Volatility makes the case sharper. In the NAM's Q1 2026 survey, 63.3% of manufacturers reported increased supply chain volatility over the past 12 months.
See it live on October 8
Reese Summers and Telka Clem from Western Computer are walking through this in a demo company on Thursday, October 8, at 1:00 PM CT: forecast, plan, promise and ask, on one item from start to finish. It's 30 minutes with Q&A. After nearly 40 years and 1,750+ implementations, we're glad to share what's ready and what still needs groundwork.