Use Case: NexusIQ – Intelligent Operations Platform

DEMO

NexusIQ - Intelligent Operations Platform

Answer complex questions about supply-chain operations in seconds by federating structured data with unstructured runbooks and incident notes

Building Blocks used:

ai.agents

data.lakehouse

Skills used

Download skills for IBM Bob →

Products Used

1 Review use case

Modern enterprises hold their most valuable operational knowledge in two completely disconnected worlds: structured records locked in databases and data warehouses, and unstructured institutional knowledge buried in runbooks, incident reports, and policy PDFs. Today, answering a single operational question — 'Why is this order delayed, and what should we do?' — requires an analyst to manually query multiple systems, search through documents, and correlate results before anyone can act. That process takes hours.

This demo collapses that time to under two seconds. It showcases a fully agentic architecture built on IBM watsonx.data as the unified query layer, with Apache Iceberg hosting all structured operational, analytics, and lakehouse data (covering orders, customers, warehouses, inventory, shipment events, and compensation policies). Unstructured content — four operational runbooks and two historical incident reports — is stored as vector embeddings in DataStax Astra DB or HCD, enabling semantic retrieval.

A lightweight Python MCP (Model Context Protocol) server bridges the AI agent to both data worlds, exposing seven tools: SQL execution, vector search, and five schema-exploration helpers. An optional Confluent Kafka pipeline streams live orders into Iceberg with sub-500 ms end-to-end latency.

The entire system is orchestrated through IBM Bob — a VS Code AI assistant — using a custom 'AI Driven Analyst' mode that understands the demo schema, enforces correct tool selection, and formats answers for a live audience. The result is a demo that illustrates, in plain English, how structured facts and unstructured guidance combine into a single actionable insight: root cause, inventory reroute option, compensation amount, and runbook citation — all in one response.

Start Here

Watch this short overview

Preview the application

Explore key screens from the demo

2 Try the interactive application

Experience the interactive demo and explore its features. A demo script is provided to help when presenting the application to partners.

Have questions, want a walkthrough or found a bug?

Contact the demo owner for assistance.

Submit a request →

3 Explore how it was built

High level overview of major demo components

Architecture

4 Download and modify with IBM Bob

Download the source code and specifications for this application to set up and run it locally. Once downloaded, open the code in IBM Bob to modify and adapt it to your own use case.

Source Code

Script Duration: 15–20 minutes | Audience: Technical stakeholders, architects, data engineers, operations & C-suite

1. Set up the environment (once)

Run demo/scripts/manual_load_iceberg_all.sql in the watsonx.data Query Workspace to create 9 Iceberg tables and 70 rows of demo data. Copy mcp-servers/query-layer/.env.example → .env and fill in Presto host/credentials and Astra DB / HCD connection details. Run python demo/scripts/reset_and_load_astra.py to embed and load the 6 runbook PDFs into the vector store.

2. Open the project in VS Code with IBM Bob

Switch Bob to the “AI Driven Analyst” mode (defined in .bob/custom_modes.yaml). The MCP server registered in .bob/mcp.json starts automatically — verify the Bob status bar shows the MCP server connected.

3. Tell the business story first (2 min)

Set the scene: “A PLATINUM customer — €250K lifetime value — has a delayed order. The root cause, the fix, and the compensation policy live in three different places. Watch this get answered in 2 seconds.”

4. Run the five demo question blocks

Q1 (structured root cause): “Why is order O-10452 delayed?” — Bob queries orders, inventory, and shipment events via SQL.
Q2 (federated): “Why is O-10452 delayed and what should we do per our runbooks?” — Bob runs SQL + vector search, cites runbooks rb-1, rb-2, rb-4 and past incident inc-2.
Q3 (customer profile): “Tell me about customer C-9001.” — Demonstrates cross-table join across tier, LTV, region, and segment.
Q4 (revenue at risk): “How much revenue is at risk from delayed orders?” — Aggregation across tiers and regions.
Q5 (real-time, optional): Start the Kafka consumer, produce a new order, then ask “Show me orders from the last 5 minutes” — the new order appears in <500 ms.

5. Close with the business value summary

Emphasise the five outcomes: 100× faster decisions, consistent policy application, proactive service, €250K LTV protected, and a platform pattern reusable across industries (healthcare, financial services, manufacturing, retail).