AI Demand Forecasting & Inventory Optimization

It is not an instant AppSource download. ECOSIRE builds the extension to fit your item catalog, locations, and planning policies, then installs it as a per-tenant extension on your Business Central environment. Typical delivery is a few weeks depending on data quality and the number of item/location combinations. We confirm a firm lead time after a short scoping call and a look at your sales and consumption history.

After go-live ECOSIRE provides support and keeps the extension compatible with BC's monthly platform updates under an agreed support window. The code uses an upgrade-safe codeunit and interface structure plus Application Insights telemetry so we catch failed forecast runs or scoring-latency issues early. Fixes, model retuning, and minor enhancements are handled as part of the support arrangement; larger new capabilities are scoped separately.

Both options are supported. Lighter statistical and seasonal models can run inside BC on a job queue schedule. For heavier machine-learning models, the extension calls an ECOSIRE-managed ML scoring gateway over the BC REST/OData v4 API and writes results back into your forecast tables. We choose the approach based on your catalog size, accuracy needs, and data-residency requirements.

It reads posted Item Ledger Entries, sales lines, and consumption history already in Business Central, so no external data warehouse is required to start. More history (ideally 18–24 months) and clean item/variant/location structure produce stronger seasonality detection. During scoping we assess your history depth and flag any data gaps that would affect forecast quality before we build.

It complements native planning rather than replacing the engine. Forecasts and dynamic safety-stock/reorder-point suggestions are written into the standard requisition and planning worksheets via event subscribers, so your team keeps the BC workflow they know. You can run it alongside existing reordering policies and adopt suggestions item group by item group, with manual override and forecast-locking for events the model can't see.

Yes. The extension tracks forecast accuracy and bias (MAPE, bias %, tracking signal) on a dedicated table and an Item Card factbox, so planners can see exactly where the model is reliable. Planners can override any suggestion and lock forecasts for promotions or new product introductions, and those overrides are preserved on the next recompute.