Demand forecasting with AI: Orpheon, NexPatch's forecasting platform.
Orpheon is NexPatch's forecasting platform for demand forecasting with AI. It turns operating data you already have - from ERP, inventory management, or sensors - into forecasts for sales, energy demand, and spare parts. NexPatch builds and operates Orpheon itself, deployed entirely behind your firewall on request, so your data never has to leave the company.
What Orpheon is
Orpheon is NexPatch's own product: a forecasting platform that turns operating data into forecasts you can act on. Instead of spreadsheets and gut feel, Orpheon works from the historical and live data already flowing through systems your company runs anyway. The platform belongs to NexPatch's private AI infrastructure: it can run entirely inside your own system boundary, in a private cloud environment or on your own hardware. That sets Orpheon apart from pure cloud forecasting tools, where operating data flows to an external provider by default. With Orpheon, it does not have to - you decide where your data and models live.
Technically, Orpheon is built on the same principles as the rest of NexPatch's private AI infrastructure: access rights are granted granularly, models run in a controlled environment, and every step of data processing is traceable. For companies with strict requirements around data protection or trade secrets - manufacturing and healthcare, for example - running behind your own firewall is often a basic precondition for using AI at all, not a nice-to-have.
Beyond the forecast calculation itself, Orpheon gives business departments a consistent interface, so purchasing, production, or maintenance teams can work directly with the results without setting up their own data analysis.
Who works with Orpheon
Orpheon is not built for data teams alone. Once it is rolled out, three groups typically use the platform, each with different access. Business departments - purchasing, production, or energy management - read forecasts and act on them. Data teams build and maintain the underlying pipelines. IT staff monitor the platform's operation inside your own infrastructure and manage access rights. This separation of roles is what makes forecasts land in day-to-day decisions, rather than sitting as an analysis result inside a single department.
Three use-case scenarios for demand forecasting with AI
In practice, Orpheon covers three recurring forecasting tasks. They differ in their data sources and target figures, but share the same technical foundation of data pipelines, model operations, and interface.
| Scenario | Typical data sources | Typical benefit |
|---|---|---|
| Sales forecasting | Order history, point-of-sale data, ERP master data | Basis for purchasing and production planning |
| Forecasting energy demand | Meter data, production plans, weather data | Reliable planning for procurement and load management |
| Forecasting spare-parts demand | Maintenance history, machine data, bills of materials | Fewer stockouts and lower inventory levels |
For sales forecasting, Orpheon reads order and sales data and calculates expected values for the coming weeks or months. Seasonal patterns, promotions, and range changes feed in as explanatory variables. Unlike a rigid spreadsheet model, Orpheon adapts as new data arrives, without a department manually reworking formulas. Sales and production planning end up sharing one number base instead of maintaining separate ones.
For manufacturers and energy suppliers, Orpheon calculates expected energy demand from production plans, historical consumption, and external factors like temperature or calendar effects. The result feeds procurement, peak-load management, and internal budget planning. Especially when energy prices are volatile, a reliable outlook helps make procurement decisions earlier and with less uncertainty, instead of buying at unfavorable terms at short notice.
For manufacturing operations, Orpheon combines maintenance history with machine data and bills of materials to estimate expected spare-parts demand per item. That reduces unplanned downtime caused by missing parts while avoiding excess inventory at the same time. Maintenance teams get a priority ranking of which parts to reorder first, instead of treating every line item the same. A deeper look with a worked calculation is on the Orpheon Forecasting page.
How Orpheon is built
Orpheon is built from three technical layers, each described on its own dedicated subpage.
Before a model can compute anything, data from source systems has to flow in, get cleaned, and land in a consistent format. Orpheon handles that through visually built data pipelines, with no traditional coding required. These pipelines can keep growing - for example when a new source system arrives or an existing data source changes - without halting the rest of model operations. Details are on the Orpheon Pipelines page.
The forecasting models themselves run as a service inside Orpheon's infrastructure. NexPatch operates this layer, monitors result quality, and swaps models out when new data or shifting patterns call for it. That includes ongoing oversight: if a forecast repeatedly drifts from what actually happens, NexPatch investigates the cause and adjusts the model or the underlying data, rather than leaving a once-trained version running unchanged.
Business departments access forecasts through a graphical interface, without writing their own queries. Results can come back as a table, a chart, or through an interface into existing systems. Access rights can be set so each department sees only the forecasts relevant to it - which matters especially in larger organizations with multiple plants or sites. How this feeds back into ERP and inventory systems is described on the Orpheon Integrations page.
From first pilot to productive operations
NexPatch applies the same approach to Orpheon as to its other private AI projects: an initial internal pilot stands within 72 hours, so business departments can work early on with real data and real forecasts. Productive rollout into day-to-day operations follows in 4 to 10 weeks, depending on data readiness and system landscape. Once live, the infrastructure runs to a general operating target of 99.9% availability - a general operating goal NexPatch pursues, not a guarantee made to any individual customer. Details on ongoing operations are on the Operations page.
Proof: how NexPatch uses Orpheon itself
NexPatch uses Orpheon internally for its own demand and capacity planning, building practical experience with the platform before it reaches customers. This self-use is explicitly not a customer reference - it is proof that NexPatch runs its own product in its own operations. Verified customer references with result figures will appear on the References page once approvals are in place. Anyone who wants a first sense of the scale of possible effects for their own operation can start with the calculator.