Skip to main content

Customer 360: M&A Consolidation 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.

This lab runs entirely through Boomi Companion

Every build step will be done with Boomi Companion — the Boomi Companion skills running in Claude Code (or your preferred AI-driven development tools) on your computer. You use the platform UI only to review and verify what Companion built. Setup is covered in the Prerequisites.

Business Situation​

Acme Corporation has just acquired BetaCorp. Acme runs on Salesforce and Marketo — clean, structured, well-governed customer data. BetaCorp ran on Dynamics 365 and SAP. The integration team has been handed an aggressive mandate: deliver a trusted, unified Customer 360 across the combined business in 90 days, so leadership can finally see overlap, cross-sell potential, revenue synergy, and churn risk across both estates.

But here is the catch the whole project turns on: BetaCorp's real customer truth is not in clean tables. The true legal entity names, parent/subsidiary relationships, contract values, renewal dates, and account context live in signed contract PDFs, free-text CRM notes, and support tickets — exactly the material a deterministic master-data engine cannot read. Before DataHub can match and survive anything, someone has to turn that unstructured mess into structured records. Today that is weeks of analyst hand-keying. In this lab, an AI agent does it instead.

Challenges​

  • The most valuable M&A truth is locked in unstructured text. True legal entity names, corporate hierarchy, and commercial terms live in contracts, CRM notes, and tickets — not in the structured fields a rules engine can read.
  • Manual hand-keying is slow and error-prone. Analysts spend weeks transcribing acquired-company contracts and notes into structured records before any consolidation can even begin.
  • A deterministic MDM engine is structurally blind to that text. DataHub matches on string similarity only; it cannot reason over a contract clause, a CRM note, or a ticket — so the hardest identity signals never reach it.
  • Ungoverned AI is not an option in an audited consolidation. Any automation that touches customer master data must be secured, rate-limited, traceable, and must defer ambiguous calls to a human — not guess.

Solution​

You will build an Unstructured Onboarding Analyst agent in Agentstudio that reads BetaCorp's unstructured customer material — contracts, CRM notes, support tickets — extracts the structured entity, hierarchy, and commercial signals locked inside, and publishes clean candidate records into DataHub. DataHub then does its native, deterministic job: match, merge, survivorship, and golden records. Companion accelerates the build, Integration onboards the structured sources, and API Management governs every action the agent takes. (In production, an Integration process can wrap the agent in an automated onboarding flow — in this lab you invoke it conversationally.)

The design crux — additive, not redundant

Match, merge, and survivorship are DataHub's native, deterministic job, and this lab is deliberately designed so the agent never enters that lane. The agent does the one thing DataHub structurally cannot do: reason over unstructured text. Boomi's own documentation states that DataHub cannot perform matching on Long Text fields — its matching is string-distance only, never semantic.

Stated plainly: the agent FEEDS DataHub; DataHub still DECIDES. The agent reads the contract, extracts the structured record, and proposes it. DataHub's engine and human stewards decide whether two customers are the same and which fields survive. The agent never declares a match, never merges, and never runs survivorship. That is what makes it additive value rather than a re-implementation of DataHub.

Use Case​

In this lab, you will build an end-to-end M&A Customer 360 consolidation across five Boomi pillars, centered on one business object — the Customer.

You will:

  1. Use Companion to turn the consolidation requirements into a recommended architecture, a DataHub Customer model, and integration skeletons.
  2. Explore the CustomerGolden360 DataHub model with three sources, three-tier match rules, and survivorship — the deterministic system of record (and build it yourself if you are in your own trial).
  3. Use Integration to onboard BetaCorp's structured Dynamics and SAP sources, normalize them to a common Customer shape, and publish them into DataHub.
  4. Build the Unstructured Onboarding Analyst agent through Companion, with governed read and write tasks and guardrails that keep it out of the match/merge lane.
  5. Walk the governed managed APIs behind the agent's actions and the MCP source that exposes them as agent tools.
  6. Run an unstructured artifact end-to-end: agent reads it, publishes a candidate, and DataHub matches and survives it into a golden record.
  7. Build the read-only Customer 360 Analyst agent that answers natural-language questions over the golden records, fusing the structured master with the unstructured contract truth.

Business Outcomes​

  • Faster, defensible integration. Companion's requirements-to-build acceleration plus an agent that eliminates weeks of manual analyst hand-keying of unstructured acquired-company data.
  • A trusted system of record from better input. DataHub builds golden records from richer signals, because the agent surfaced identity and relationship truth that was previously locked in PDFs and notes.
  • Reaches the data that mattered most. The hardest M&A truth — true legal entity, corporate hierarchy, contract terms — lives in unstructured form; the agent unlocks exactly what the rules engine cannot see.
  • Governed, auditable AI. Every action the agent takes flows through a managed API, so its contribution is secured, rate-limited, and traceable; DataHub stewardship remains the system of record.

Key Performance Indicators (KPIs)​

  • Unstructured Records Onboarded: Volume of contracts, notes, and tickets turned into structured DataHub candidate records.
  • Extraction Confidence / Auto-vs-Steward Rate: Percentage of artifacts the agent structured confidently versus routed to a human steward.
  • Customer Records Consolidated: Golden records produced by DataHub from the combined structured and agent-sourced input.
  • Time-to-Trusted-View: Elapsed time from requirements to a governed, queryable Customer 360.

Workshop Structure​

This flagship workshop is organized into seven parts spanning five platform pillars. Each part builds on the last, so work through them in order.

Part 1: Architecture with Companion​

Hand Companion the consolidation requirements and let it propose the integration architecture, the DataHub Customer model, and the integration skeletons that make a 90-day timeline credible.

  • Requirements-to-build acceleration.
  • A recommended Customer model and source layout you carry into Part 2.

Part 2: The DataHub Golden Customer Record​

Use Companion to introspect the pre-built CustomerGolden360 model in the shared account — its fields, three sources, three-tier match rules, and survivorship — and query what the golden data looks like. (Building it in your own trial? The creation prompts are included and flagged for you.)

  • The deterministic MDM lane — the agent stays out of it.
  • Match rules engineered to keep near-duplicates (Stark vs. Starke) distinct.

Part 3: Onboard BetaCorp Structured Sources​

Use Integration to onboard the structured Dynamics and SAP feeds, normalize them to the common Customer shape, and publish them into DataHub.

  • Flat-file profiles, a cross-reference join, and a normalization map.
  • A live publish into the BETACORP source.

Part 4: Govern the Agent's Tools with API Management​

Set up the governed surface the agent will use — register an application, subscribe to the API product, and review the MCP source that exposes the action endpoints as tools.

  • Security, rate limits, and observability on every agent action.
  • HTTP-transport MCP source and the query-param OpenAPI-tool note.

Part 5: Build the Unstructured Onboarding Analyst Agent​

Build the agent entirely through Companion prompts — six read/reason tasks, two governed write tasks, and Control Tower guardrails — then review the result in Agentstudio.

  • Reasoning over contracts, notes, and tickets DataHub cannot read.
  • Guardrails that propose and defer, but never match or merge.

Part 6: Run It End-to-End​

Run an unstructured artifact through the agent into DataHub and watch the two layers cooperate — agent reads, DataHub decides.

  • Six engineered outcomes: three clean matches, a net-new insert, a subsidiary the contract reveals, and a governed steward exception.

Part 7: Talk to Your Customer 360​

Give it a voice: a read-only Customer 360 Analyst agent that answers natural-language questions over the consolidated golden records and fuses the structured master with the unstructured contract truth in a single answer.

  • Provenance-aware 360 lookups (structured record + what the contract revealed).
  • Portfolio questions: overlap, revenue synergy, record counts.
Ready to Get Started?

Let's build your Customer 360! Click Next below to begin with the prerequisites.