Skip to main content

Part 7: Talk to Your Customer 360

You've built the trusted golden records. Now put a conversational face on them. In this part you build a second agent — a read-only Customer 360 Analyst — that answers questions over the consolidated golden records and, uniquely, fuses the structured master record with the unstructured contract truth in a single answer.

This is a different role from the Unstructured Onboarding Analyst you built in Part 5. That agent feeds DataHub (it reads messy documents and proposes structured candidates). This one reads the finished result — and reaches back into the original documents when you ask what a contract revealed.

Like Part 5, you first have Companion draft a plan from a few simple requirements, then build the agent from that plan and deploy it — then open Agentstudio to review it. The one exception is the DataHub Query tool, which is pre-built for you (see the next section for why).

Read-only and additive

The Customer 360 Analyst never creates, edits, or merges records. It queries the golden records and reports what they say, with provenance. DataHub still owns match, merge, and survivorship.

What You'll Build​

  • A new Customer 360 Analyst agent (separate from the onboarding agent).
  • A DataHub Query tool that reads the consolidated CustomerGolden360 golden records — pre-built for the workshop (the one component not created by the build prompt).
  • The read-only unstructured tools (read-artifact, extract-entity, extract-hierarchy, extract-commercial) so the agent can pull the original contract/notes and fuse them with the structured record.
  • Read-only guardrails so the agent stays in its lane.

The DataHub Query Tool — the One Pre-Built Step​

Everything else in this part is driven by Companion from your plan. The DataHub Query tool is the exception, and here is why.

Known issue: DataHub Query tools must be built with the progressive flow

Creating a DataHub (hub) tool with a single API call currently produces a tool that looks ACTIVE but fails at query time — the field definitions and universe binding do not resolve correctly from one shot. The reliable method is the platform's progressive build flow (create a blank hub tool, read the authoritative field list from the universe endpoint, PUT those fields, then install). Because a Companion "build the tool in one prompt" would hit this bug, the DataHub Query tool is pre-built for the workshop and you attach it by name. (The full flow is documented in the agent-management skill's tool API reference.)

  1. Confirm the pre-built DataHub Query tool exists. In Agentstudio it is registered against the CustomerGolden360 universe in the Customer360MA repository under the name:

    Customer360 MA CustomerGolden360 Query
    Trial-account builders: build this tool yourself first

    If you are reproducing this lab in your own trial (see the appendix), the shared pre-built tool is not in your account — build your own DataHub Query tool over your CustomerGolden360 universe using the progressive build flow (create blank → read fields from the universe endpoint → PUT fields → install), per the agent-management skill. Give it a name you'll recognize (e.g. [builderInitials] CustomerGolden360 Query) and substitute that name in the build prompt below.

Plan the Analyst​

We've built and consolidated the trusted golden records; now we'll put a conversational, read-only face on them. As in Part 5, we start by having Companion make a plan.

Our analyst agent needs to be able to do the following things:

  1. Answer natural-language questions over the consolidated Acme/BetaCorp golden records in DataHub — a customer's full 360 (identity, parent/subsidiary hierarchy, commercial terms, contributing sources, and provenance) as well as portfolio questions like Acme/BetaCorp overlap, cross-sell, total revenue synergy, and churn risk.

  2. When asked, reach back into the original BetaCorp contracts and notes and fuse that unstructured truth with the structured golden record in a single answer.

  3. Stay strictly read-only. Always ground answers in the golden record and the source documents, never invent data, and never decide whether two customers are the same — that stays DataHub's job.

  4. Use the pre-built DataHub Query tool (attach it by name, do not create a new one) plus the four read-only unstructured tools from the Customer 360 MA API MCP source; do not attach either write tool.

  5. Offer a few conversation starters that show it off — a full 360 lookup, a portfolio/synergy question, and a simple record count.

Copy this text, and ask our Companion session to make a plan to execute this.

Build the Analyst​

Now that we have a plan, we can build our agent. Instruct Companion to build the analyst we've planned.

  1. In Companion:

    Based on the plan we just developed, implement the read-only analyst agent in Agent Studio using the agent-management skill. Name it [builderInitials] Customer 360 Analyst, and attach the pre-built DataHub Query tool "Customer360 MA CustomerGolden360 Query" by name rather than creating a new DataHub tool.

  2. Confirm Companion reports the agent created with: the Professional voice and objective; the DataHub Query tool attached by name; the four read-only unstructured tools attached and neither write tool; two guardrails; and three conversation starters.

    Attach the DataHub tool, don't recreate it

    The plan tells Companion to attach the existing Customer360 MA CustomerGolden360 Query tool by name. If Companion instead tries to create a new DataHub tool in one shot, it will hit the known issue above and the tool will fail at query time — so if the build reports a new DataHub tool rather than attaching the pre-built one, remove it and attach the pre-built tool by name.

    Ask questions; don't say "combine"

    The no-merge guardrail is deliberately strict. Phrase prompts as questions — "show me the 360 for Hooli, including what the contract reveals" — rather than "combine these records." The word "combine" can trip the guardrail before any tool runs.

Deploy the Analyst​

  1. Ask Companion to package and deploy the agent:

    Package and deploy my [builderInitials] Customer 360 Analyst agent, and confirm the deployment reaches ACTIVE.

  2. Confirm the deployment reaches ACTIVE.

    Agent activation

    An agent activates by creating a package and then a deployment. Once the deployment is ACTIVE on the Agent Garden runtime, the analyst is invokable conversationally.

Review in Agentstudio​

Companion did the building; now use the platform UI to review the result. This step is review only — do not edit the agent in the UI.

  1. In the platform, open Agentstudio → Agent Garden and open your [builderInitials] Customer 360 Analyst agent.

  2. Verify the tools: the DataHub Query tool against the CustomerGolden360 universe plus the four read-only unstructured tools — and confirm neither write tool is attached.

  3. Verify the two guardrails — No Record Writes and No Match / Merge / Survivorship Decisions — and the three conversation starters.

  4. Verify the deployment shows ACTIVE.

Test the Analyst​

  1. Try each starter against the deployed agent:

    Show me the Customer 360 for Hooli, including what the signed contract reveals about any subsidiaries our master record is missing.

    The agent calls both tool types — it reads the golden record and the Hooli contract — and fuses them: the structured master (Hooli, Inc.) plus the subsidiary Pied Piper Data Services, LLC ($312,000 / NET60) that existed only in the contract.

    What is our total revenue synergy across the consolidated customer base, and how many customers overlap between Acme and BetaCorp?

    Returns $1,197,000 in revenue synergy and an overlap of 4 customers, grounded in the golden records.

    How many total customer golden records are there?

    Returns 8 consolidated golden records.

    If a query errors, just ask again

    The DataHub query occasionally returns a transient execution error on the first call. Re-ask the question and it resolves — the agent will not invent numbers when a query fails.

What You Now Have​

  • A read-only Customer 360 Analyst that answers natural-language questions over the consolidated golden records and fuses structured master data with unstructured contract truth on demand, always grounded and always provenance-aware — built from a plan and a simple Companion build over a pre-built DataHub Query tool.
Congratulations!

You've completed the full Customer 360: M&A Consolidation lab — and given it a voice. Across the entire Boomi platform, driven end-to-end through Companion, you drafted the architecture, explored a deterministic DataHub golden record, onboarded structured sources with Integration, stood up a guardrailed Unstructured Onboarding Analyst that reached the truth locked in contracts and notes, governed its tools through API Management, ran the consolidation end-to-end, and finally built a conversational analyst that fuses the structured master with the unstructured source truth in a single answer. The agent feeds and reads; DataHub decides; a human steward owns the ambiguous case — faster, defensible integration, a trusted system of record, and governed, auditable AI that leadership can simply ask!