Project snapshot
The client’s supply chain management function relied on a patchwork of SaaS tools and ERP systems, each holding a different piece of the data. Intellias built a data product to bring forecasting and planning data out of a third-party ML-based system, clean it, and make it available across the business without duplication. The result is an AI-ready data platform for supply chain operations.
The initiative is the first of a larger data strategy the client is rolling out company-wide: domain-owned data products, each with a clear owner, designed to eventually connect with each other and support enterprise AI use. It is also an example of AI-ready data architecture for enterprises that want to scale beyond isolated pilots.
Business challenge
Supply chain data had historically been treated as a byproduct of transactional systems rather than a managed asset. Technology decisions preceded data architecture, resulting in domain knowledge scattered across siloed platforms, high coordination overhead between teams, and repeated consolidation attempts that failed because they weren’t built around the data domain. The company lacked a unified data foundation and a coherent data platform strategy.
The client was already moving toward a Snowflake-based data-sharing model to reduce duplicated consolidation work. What was missing was a standardized way to expose the forecasting engine data to the rest of the organization, feed predictions back into transactional processes such as order proposals, and provide a clean, secure foundation for AI and analytics use.
The main challenge was organizational rather than technical: no data owners were assigned. The team had to identify the relevant datasets and stakeholders from scratch before building any pipelines.
Solution
Intellias built the data product on Databricks, using a zero-copy approach to ingest data from the forecasting engine rather than duplicating it into a separate store. L0/L1/L2 data layers were defined from the outset, separating raw, cleaned, and business-ready data. Together, these choices form an AI-ready data architecture that keeps data governed and reusable.
The operational data model covers forecasts, SKU-level predictions, and order proposals, supported by a transformation and enrichment layer. The serving layer supports multiple consumption methods, including API, file share, Confluent streams, and direct Snowflake access. This enables data streaming for Snowflake and other downstream platforms, and it accommodates different types of consumers. As a set of AI-ready data solutions for supply chain management, it serves both operational reporting and AI/analytics use cases.
The project is delivered in two stages. Stage one, now live, moves data from the third-party forecasting system into the client’s internal systems. Stage two, underway and targeted for early next year, sends data back to the forecasting engine to support its order proposal calculations.
Business outcomes
The first data flow is live, providing a working example of a governed data product with no duplication and clear ownership. Onboarding downstream consumers is in progress, which will determine the specific business value delivered. Beyond the immediate integration, the project delivered:
- A reference implementation for the client’s broader data strategy. It serves as a blueprint for the eleven other data products already planned, replacing a trial-and-error approach with a proven pattern and a repeatable data foundation.
- A governed foundation for AI readiness. With defined data layers, clear ownership, and a serving layer built for both operational reporting and AI/analytics consumers, the project demonstrates what an AI-ready data platform looks like in practice — the foundation that determines how quickly the client can move from isolated AI experiments to models it can trust in production.
- Support for regulatory compliance. The architecture is designed to interconnect with the systems supporting EU regulatory requirements, so the same governance work also strengthens the client’s compliance posture.
Together, these outcomes have set the direction for how the client intends to manage data across the business going forward: owned, governed, and built for AI from the start.