Skip to main content

Sourcing Agent Lab

This is an Instructor-Led Lab

This lab is designed for in-person workshop events led by a Boomi instructor. Participants must be provisioned with access to a dedicated Boomi platform account during the workshop. This lab is not intended for self-led learning.

Looking for Self-led learning?

For self-paced learning, check out the Customer Support Triage (Extended) Lab.

Draft Lab

This lab is currently in draft form, derived from the Boomi World 2026 Sourcing Agent research package. Screenshots and exact UI click paths will be added after the reference build is validated in a trial tenant. Tool names, agent profile guidance, and integration patterns are stable.

Business Situation​

Welcome to Northwind Components, a mid-market industrial distributor whose sourcing team negotiates against a moving target. Supplier price sheets refresh on different cadences, public benchmark indices shift weekly, and the contract review meeting always lands on the day the data is the most stale. By the time anyone notices a vendor has drifted ten percent above market, the next quarterly purchase order has already been cut.

Northwind does not need another dashboard. They need a colleague who watches prices while the team sleeps, flags drift before it shows up in a renewal, and writes the recommendation down in a place the buyer trusts. In this lab you will build that colleague — the Sourcing Agent — and wire it into the buyer's request flow so that every sourcing event ends with a defensible, ranked vendor shortlist instead of a spreadsheet hunt.

Challenges​

  • Price drift goes unnoticed: Supplier rates and public benchmarks refresh asynchronously, so vendor overcharges of 5–10% routinely slip through to the next purchase order
  • Manual benchmarking is too slow: A buyer pulling internal price history, calling the benchmark service, and reviewing vendor performance ratings burns hours per sourcing event
  • Recommendations are unstructured: Vendor decisions live in email threads and spreadsheets with no consistent rationale, making renewals and audits painful
  • Approvals stall without context: When a buyer-initiated request reaches review, the approver lacks a single artifact tying ranked vendors to deviation severity and historical performance
  • Off-hours drift has no owner: No one is watching sourcing events that close without a recommendation, so opportunities to renegotiate or switch vendors are missed entirely

Solution​

A Sourcing Agent that optimizes spend by benchmarking supplier prices against internal history and external market rates, then recommends vendors with a defensible rationale. The agent is invoked when a buyer initiates a sourcing event and writes a structured ranked-vendor shortlist into the buyer's review workflow. MEDIUM and HIGH severity recommendations land on a shared Boomi Flow approval dashboard, where a human reviewer approves or rejects each one before it drives a PO.

Use Case​

In this lab, you will build a Sourcing Agent on the Boomi platform that automates supplier benchmarking and vendor recommendation. The agent will:

  1. Pull internal price history: Query the last 12 months of ERP purchase data for the relevant vendors and SKUs
  2. Fetch external market benchmarks: Call an industry rate index for the same SKU class to establish current market norms
  3. Look up approved vendors: Retrieve the vendor directory with active contract terms and payment windows
  4. Surface vendor performance: Read on-time delivery rate, defect rate, and responsiveness scores per vendor
  5. Write a ranked recommendation: Generate a structured shortlist with severity bands (LOW / MEDIUM / HIGH) and a recommended action (HOLD / RENEGOTIATE / SWITCH_VENDOR)

The buyer-initiated trigger calls the agent synchronously and routes the resulting recommendation_id to a buyer review surface for approval. On approval, downstream PO creation picks up the recommended vendor.

Trial-Tenant Safe Path

This lab uses API Tools as the core agent surface so it works on every Boomi trial tenant today. Reference data (vendor list, internal price history, vendor ratings) is bootstrapped by an idempotent ephemeral seed process — no Master Data Hub seeding required. A future-state appendix shows how to swap the flat-file backing for DataHub once trial-tenant DataHub seeding is solved.

Business Outcomes​

  • Spend optimization: Catch vendor price drift the moment a sourcing event opens, rather than discovering it during a quarterly review
  • Decision auditability: Every recommendation is persisted as a structured artifact with ranked vendors, deviation severity, and rationale — defensible at renewal and audit time
  • Buyer productivity: Eliminate the manual benchmarking lap so buyers spend their time on negotiation and supplier relationships, not data assembly

Key Performance Indicators (KPIs)​

  • Average Price Deviation vs Benchmark: Track the percentage gap between supplier prices and public market benchmarks across active sourcing events
  • Time to Defensible Recommendation: Measure how long it takes from sourcing-event open to a structured ranked-vendor shortlist landing in the buyer's review queue
  • Recommendation Approval Rate: Monitor the percentage of agent-generated recommendations that buyers approve without modification, indicating trust and rationale quality

Workshop Structure​

This hands-on lab is divided into three parts, guiding you through the complete implementation of a Sourcing Agent. The full lab takes 4–6 hours and is delivered in three 2-hour blocks.

Part 1: Create the Five Sourcing Tools​

Build the five Boomi processes that the agent will call as tools. You'll start with an idempotent ephemeral seed process that loads mock vendor, price-history, and rating data into process-local stores, then expose each tool through API Tools so the agent can call it by name. By the end of Part 1 you will have five deployed tools — Internal Price History, External Market Benchmark, Vendor Directory, Vendor Rating Lookup, and Recommendation Output Writer.

Part 2: Build the Agent with AI​

Use Boomi's Build with AI assistant to generate a defensible Sourcing Agent profile in a single step. You'll paste a one-paragraph problem statement, edit the generated profile to match Boomi voice, attach the five tools to the agent's tool surface, and version the agent. By the end of Part 2 you will have a deployed agent that can rank vendors and write structured recommendations.

Part 3: Integrate the Agent​

Wire the agent into a human-in-the-loop (HITL) review surface: add a sixth task that publishes MEDIUM and HIGH severity recommendations to a Boomi Event Streams topic, attach the pre-built Publish Recommendation to HITL Stream Integration tool, and exercise the end-to-end flow on a shared Boomi Flow approval dashboard. Approve or reject a row and watch a tombstone retire it from the dashboard while a DECISION_MADE event lands on the decisions topic for downstream consumers.

Ready to Get Started?

Let's build your Sourcing Agent! Click Next below to begin with the prerequisites.