Summary:
We are excited to announce the launch and continued rollout of Product Intelligence Engine or PIE. This milestone represents the culmination of extensive collaboration, iteration, and innovation across teams, resulting in the successful rollout of 6 category code predictions across 583 vendors in 2 regions.
Question: What is Product Intelligence Engine and how does it work?
Answer:
PIE predicts the fields required for SKU creation using vendor-provided descriptions and other supporting contextual data. It relies on human in the loop at scale with tools developed for this. Our platform leverages three complementary approaches:
PIE - Machine learning-powered semantic retrieval with continuous Human-in-the-Loop feedback through our AI training and data workbench, PIE Studio.
PIE Reasoning - An LLM-powered categorization agent that applies reasoning to determine the most accurate category assignments.
PIE Hybrid - A best-of approach combining Auto Indexing, PIE Retrieval, an LLM Judge, and PIE Reasoning to maximize prediction quality.
RESULTS: In this space, the cost of an incorrect prediction can exceed the cost of a manual process, making precision critical. After two rounds of training and refinement, PIE now delivers strong prediction accuracy across vendors:
HPE AMER: 95%
HPE EMEA: 91%
HPI AMER: 86
HPI EMEA: 90%
Lenovo AMER: 89%
Importantly, catalogs that previously took weeks to onboard can now be onboarded in hours, with PIE and categorization automation playing a major role in enabling that transformation.
WHAT’S NEXT: The goal is to expand these 6 category code predictions across all vendors and all regions. Ingram Micro is actively evaluating additional fields that can benefit from the platform and capabilities it has built. On the horizon are NASA SEWP VI, Digital Capacity and Renewal Flags.