Demand forecasting in Dynamics 365 F&O,
checked against the order book every day
Lleverage is an AI agent for demand forecasting in Dynamics 365 Supply Chain Management, part of D365 F&O. It reads your sales invoices, open quotations and the forecasts customers send by email, shows where they part from the authorised demand forecast, and replans production when an order, a machine or a delivery moves. Forecast changes wait for your planner.
How does D365 demand forecasting work with an AI agent?
Supply Chain Management calculates a statistical forecast from history, and master planning turns the authorised forecast into planned orders. Between two runs, demand keeps arriving in the inbox. The agent covers that part in 4 steps, and keeps the forecast and the plan in F&O.
Read what customers bought and what is coming
Sales invoices per customer and released product, open sales orders and sales quotations, company by company, because one F&O environment often holds several legal entities. Orders our order agent creates in F&O count from the minute they exist.
Catch the demand that never reaches the ERP
A customer mails next quarter’s forecast as a spreadsheet, a distributor cancels a call-off by mail. The statistical baseline sees none of it until the sales show up as invoices.
Set it against the authorised forecast
Per item, customer and period, the agent compares the signals with the demand forecast your planners authorised. A difference worth acting on goes to the planner with the invoices, quotations and mails behind it, and nothing changes in the forecast until the planner agrees.
Replan production orders on your resources
Jobs go onto resources that can take them, following the route’s operations, your changeovers and minimum runs, with components checked against open purchase orders. A new date a customer already has waits for a person. Forecast lines and production order dates are written through a custom integration our engineers set up.
What changes when the forecast in D365 keeps up?
When a customer’s forecast, a machine or a delivery moves, the plan for Dynamics 365 F&O is rebuilt around it, so the floor works from a schedule that fits this week. Trade-offs and promised dates stay with your planners.
Learn moreExample data
Replanning a week in Dynamics 365 F&O around a late delivery.
A steel delivery slips two days on a week that was already full. This is what the agent does with the schedule.
Example data
You decide which forecast and plan changes need a planner
Every constraint has a setting. A new production order date that keeps every promise goes into F&O on its own, a change to the demand forecast waits for your planner, and so does any trade-off between two customers, written out.
Example data
D365 demand forecasting and Demand planning or Lleverage: which fits when?
Supply Chain Management has two ways to forecast: the demand forecasting built into it, which sends history to Azure Machine Learning for a statistical baseline, and Demand planning, which Microsoft calls its next-generation app. We read Microsoft Learn and the Lleverage docs on 18 November 2026.
| D365 demand forecasting and Demand planning | Lleverage | |
|---|---|---|
| What it produces | A statistical baseline forecast from historical transactions. Microsoft’s experiments pick the best fit among five time series methods, and the authorised forecast feeds master planning, which creates planned orders. | yes: a forecast per customer and item that reads the order book, set against the one you authorised, and a production week that still fits your resources. |
| Statistical models | Demand planning tunes AI parameters automatically, lets AI choose the forecast model or takes your own Azure Machine Learning model, and lets planners edit at product group level and see the effect per SKU. | Builds on invoices, orders and what customers tell you. For statistical modelling across thousands of SKUs, we think Demand planning is the stronger tool. |
| Demand news from customers | Demand planning takes external signals such as promotions or stockouts as imported data. In F&O, a customer’s own forecast can be entered as a customer forecast, and master planning can include it in the overall forecast. | yes: reads the customer’s forecast spreadsheet and the cancelled call-off as they arrive by mail, and shows which forecast lines they touch. |
| Changing the forecast | Manual adjustments must be authorised before planning uses them, and authorising a chosen period lets you freeze specific buckets. Unless the box that saves manual adjustments is selected, they are discarded after authorisation. | yes: proposes each change with the invoices, quotations and mails behind it. Your planner approves it on a card in Teams or Slack, and the agent carries on with the answer. |
| Capacity | Planning Optimization schedules with finite capacity once it is switched on globally, per master plan and per resource. Microsoft advises against finite capacity where capacity changes with requirements, such as with shifts. | yes: resources, shifts, changeovers and minimum runs are constraints the agent never breaks to make room. A week that does not fit comes back with the conflict named. |
| When the week changes | Each master planning run reduces the forecast by actual demand with the reduction method you choose, and creates planned orders for what is left. | yes: a late delivery is replanned when the supplier’s mail arrives, without waiting for the next run. A change that keeps every promise goes through, and a trade-off between two customers waits for your planner. |
| Working together | Demand planning has in-product comments, restorable forecast versions and in-context communication through Microsoft Teams. Each user in production needs a Demand planning licence. | yes: your planner can ask the agent in Teams or Slack, or call it, and gets an answer from the live F&O records. |
| Getting better at the job | The built-in forecasting measures forecast accuracy and lets you remove outliers from historical data. | yes: your planners rate the agent’s work, it drafts a change to its Skill from those ratings, and the change goes live once a person approves it. |
| Fits best when | History predicts next season well, the forecast is reviewed in monthly cycles, and you plan many SKUs at product group level. | Demand moves on news that comes in by mail, and your planners spend their days reconciling it with the forecast and the production week. |
We think the authorised demand forecast belongs in Supply Chain Management, and if history predicts your demand well, Microsoft’s forecasting or Demand planning may be all you need. We think Lleverage fits when the forecast drifts from what customers are telling your sales team. More on demand forecasting for manufacturers.
Our engineers set it up with your planners, and it takes on more of the week as it learns.
Forward deployed engineers connect Dynamics 365 F&O and build the first plan beside your planners, so the forecast checks and the replanning come off the desk within a few weeks. From there it automates gradually: a correction becomes a change to the Skill once a person approves it.
They work inside your operation, starting in the legal entity where planning begins, and turn the rules in your planner’s head into constraints with your team. You set the confidence line.
The ERP and MES connections, the capacities, the changeover rules and the log are yours to configure. Our documentation and support sit behind your team, and an engineer is a call away whenever you want one.
Talk to us about either ↗Frequently asked questions
Does Dynamics 365 have demand forecasting?
Yes, two kinds. Supply Chain Management has built-in demand forecasting, which sends historical transactions to Azure Machine Learning for a statistical baseline that planners adjust and authorise. Demand planning is a separate app with automatic model tuning, what-if scenarios and its own user licence. Both feed master planning in F&O, which is where the agent picks up the work.
What is the difference between demand forecasting and Demand planning in D365?
Demand forecasting is built into Supply Chain Management and runs on Azure Machine Learning experiments. Demand planning is Microsoft’s newer app for collaborative planning, with aggregation by product group, external signals such as promotions, and forecast versions. Microsoft itself suggests Demand planning for a better forecasting experience. The agent works next to either one.
Does the agent replace Demand planning or master planning?
No. The authorised forecast stays in Supply Chain Management, and master planning keeps creating planned orders from it. The agent takes the daily work around them: reading the forecasts customers mail in, finding the invoices and quotations that contradict a forecast line, and moving production orders when a delivery slips. A forecast change waits for your planner’s approval.
Which records does the agent read and write in Dynamics 365 F&O?
Through Lleverage’s F&O connection it reads customers, released products, sales invoices, sales orders, quotations and purchase orders, and picks the legal entity first. Writing demand forecast lines and production order dates is set up by our engineers as a custom integration on your environment. More on how Lleverage connects.
How long before the agent works on our forecast?
The first process typically runs supervised within a few weeks. Forward deployed engineers connect F&O with your IT team, signed in with a Microsoft account or an app registration, and turn your planners’ rules into constraints. The agent proposes forecast changes for approval first, and takes on more once your planners trust its suggestions.
The rest of the back office.
↗
Quote to orderQuote & SellQuotes drafted, orders entered and confirmed the moment they land.
↗
Purchase to paySource & ProcureInvoices, POs and confirmations, handled against the order.
↗
Order to deliveryDeliver & SupportCustomers answered with full context, in their own channel.
↗
AP & ARPay & CollectInvoices matched, real variances flagged, the rest untouched.
↗
Master dataGovern & EnableMaster data kept clean and in sync across every system.