DATA INTEGRATED. DECISIONS ELEVATED.
Vertex AI Agent Development
Overview
The client needed a faster, more consistent way to understand why service levels moved, where in-stock issues originated, and what to do next — without stitching together answers from multiple tools and teams. D&I Integrators designed and built a Vertex AI–powered agent platform that gives the retailer a single conversational interface across service-level explainability, in-stock diagnostics, KPI context, and recommendations.
Instead of separate dashboards and one-off analyses, users can now ask a question in plain language and get a grounded, evidence-based answer — backed by approved metrics, charts, and tables — in seconds.
The Challenge
Explaining service-level performance at the organization meant pulling from several disconnected sources: target-service-level drivers, in-stock root-cause data, and KPI reporting. Getting a complete, trustworthy answer required manual cross-referencing across teams and tools, which slowed decision-making and made it hard to apply the same rigor consistently across products, distribution centers, vendors, and stores.
the solution
D&I Integrators architected a multi-agent system on Vertex AI, orchestrated to normalize any incoming question, route it to the right specialist, and return a single reconciled answer grounded in approved data sources.
Initial scope: a deployment path from Snowflake to GCP/Cloud Run and Vertex AI, with persona-scoped summaries as the first production use case.
01
Orchestrator + Routing
A platform-agnostic orchestrator agent normalizes incoming prompts, determines which specialist path a question belongs to, and coordinates evidence retrieval across the system acting as the single entry point for every user interaction.
03
Evidence-Based Answers
Rather than surfacing raw data, the platform reconciles target-service-level drivers, in-stock root causes, and KPI context into a single response complete with findings, implications, recommendations, and suggested follow-up prompts.
02
Specialist Sub-Agents
Purpose-built sub-agents were designed for the core areas of the operating model: Summary, Target-Service-Level (TSL) Explainability, In-Stock Diagnostics, Text2SQL/KPI, and Reasoning/Synthesis each focused on a specific type of question and data source.
04
Persona-Scoped Deployment
Each user's saved scope — by persona, product category, distribution center, vendor, and store — is applied automatically, with a path to Entra SSO authentication for secure, role-appropriate access.
Why It Matters
By unifying service-level explainability, in-stock diagnostics, and KPI context into one conversational interface, care and supply chain teams can move from "where do I look?" to "what should I do?" faster — with every answer traceable back to approved metrics and evidence rather than ad hoc analysis.