Skip to main content

Customer Support Triage 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.

Business Situation​

Acme, Inc. is a fast-growing company known for its innovative line of high-tech gadgets and renewable energy products. While Acme prides itself on quality products, its customer support department is struggling to keep up with the company's rapid growth and increasingly complex customer inquiries.

When a customer submits a support case—whether it's a product question, order status inquiry, address change, or troubleshooting request—support agents must manually search through multiple disconnected systems: the product catalog for technical specifications, the order management system for shipment tracking, the CRM for customer history, and the knowledge base for troubleshooting steps. They then copy and paste fragments of information from each system to formulate responses. This manual, time-consuming process creates a significant bottleneck in the support workflow.

The situation is further complicated by customers who ask multiple questions in a single message—for example, inquiring about an order's status AND requesting an address update in the same email. Traditional automation fails here, as rigid email parsers designed to extract a single piece of information (like an order number) completely miss additional requests buried in natural language.

Challenges​

  • Time-consuming multi-system research: Agents spend excessive time switching between product catalogs, order systems, CRMs, and knowledge bases to gather information
  • Multi-intent inquiries break automation: Customers often ask multiple questions in one message, but traditional rigid automations can only handle one task at a time
  • Inconsistent responses: Manual copy-paste workflows from multiple systems lead to errors, missing information, and variations in response quality
  • Natural language variations cause failures: Automated systems fail when customers write "#12345" instead of "Order Number: 12345"
  • Long case resolution times: Manual processes and failed automations delay customer responses
  • Declining customer satisfaction: Slow, incomplete responses impact brand loyalty and repeat business
  • Overwhelmed support team: Growing ticket volume and complexity exceed team capacity

Solution​

Build an AI agent that acts as a skilled Customer Support Specialist—not just an automation that performs a single task, but a digital colleague that can handle product questions, order inquiries, address updates, and troubleshooting requests in a single conversation. The agent autonomously researches across multiple systems, synthesizes information, and generates comprehensive responses that address all of a customer's concerns.

Use Case​

In this lab, you'll build an AI Agent using the principles of Architecting for Agents—evolving from the traditional Integration Mindset (building rigid, task-focused automations) to the Agentic Mindset (designing role-based digital colleagues).

Your Customer Support Agent will:

  1. Understand multiple intents in natural language—parsing questions like "What's the status of order #12345? Also, my new address is 456 Main St."
  2. Research across multiple systems intelligently—querying product catalogs, order databases, and knowledge bases based on context
  3. Handle flexible data variations—recognizing "#12345," "ORD-12345," or just "12345" as the same order
  4. Synthesize comprehensive responses—combining information from multiple sources into one complete answer that addresses ALL customer concerns
  5. Update customer records when requested—modifying address, phone, or email information with proper governance
  6. Generate professional responses—creating human-like replies that include troubleshooting steps, order updates, and contextual information
  7. Document resolutions in the CRM—writing proposed responses as internal case comments for human review
  8. Operate autonomously in business processes—activating automatically when new support cases arrive, working 24/7 without human intervention
The Multi-Intent Advantage

This lab specifically demonstrates the scenario from Architecting for Agents (pages 8-9): A customer emails "Hey, my order #12345 hasn't arrived. The tracking link says 'pending' but it's been a week. Can you help? Also, my new address is 456 Main St, in case it hasn't shipped yet."

Traditional Task-Based Automation: Would fail to parse "#12345," miss the address update entirely, and generate a useless generic template.

Your Role-Based Agent: Will identify both intents (order status + address update), check the order, update the address, explain that the order is already in transit to the old address, and provide a complete, contextualized response.

This is the power of designing for roles, not just tasks!

Business Outcomes​

  • Reduce average case handling time by 60-80% through automated research and draft generation
  • Improve response consistency and quality with AI-generated answers that address all customer concerns
  • Scale support capacity 10x without adding headcount by handling routine inquiries autonomously
  • Increase customer satisfaction through faster, more comprehensive responses available 24/7

Key Performance Indicators (KPIs)​

  • Average Case Resolution Time: Measure the time from case creation to resolution (target: reduce by 60%+)
  • First-Response Time: Track how quickly customers receive their initial response (target: under 5 minutes, 24/7)
  • Customer Satisfaction Score (CSAT): Monitor customer feedback on support interactions (target: 4.5+ / 5.0)
  • Multi-Intent Resolution Rate: Percentage of cases where the agent successfully addressed multiple customer concerns in one interaction (target: 85%+)
  • Human Escalation Rate: Track what percentage of cases require human intervention (target: under 20%)

Workshop Structure​

This hands-on workshop is divided into three parts that progressively build your understanding of Architecting for Agents:

Part 1: Create Tools for Your Customer Support Agent​

Build five flexible tools that demonstrate role-based design principles:

  • Get Product Information - With optional parameters for flexible queries (product code, category, compatibility checks)
  • Query Order Status - Handles multiple search methods and natural language variations
  • Update Customer Information - Modifies contact details with governance controls
  • Search Knowledge Base - Finds troubleshooting articles and documentation
  • Update Case with Comment - Writes resolutions back to the CRM

You'll learn how optional parameters and clear descriptions give agents the flexibility to use tools in multiple ways—unlike traditional integrations that require rigid, predefined inputs.

Part 2: Build Your Customer Support Agent​

Design and configure your Agentstudio agent as a Customer Support Specialist role with:

  • Role definition that enables multi-intent handling
  • Tasks that group tools by capability (research, update, respond, document)
  • Conversation starters that demonstrate handling multiple concerns in one message
  • Guardrails for safe operation (denying financial decisions, protecting sensitive data)
  • Instructions that implement governance patterns (confirmation, transparency, escalation)

You'll learn how to design for a role (Customer Support Specialist) instead of a task (look up product code)—enabling your agent to solve more than one problem, just like a human employee.

Part 3: Embed Your Agent into Business Processes​

Complete the transformation by integrating your agent into an event-driven Boomi process:

  • Understand the Integration Layer vs Reasoning Layer architecture
  • See how traditional integrations and agentic systems work together (complementary, not conflicting)
  • Experience event-driven activation where the agent is triggered by new support cases
  • Observe the agent synthesizing information from multiple systems autonomously
  • Review the complete end-to-end flow from case creation to resolution documentation

You'll learn how to apply agents at the right architectural layer—as the "architect, monitor, or governor" of high-performance data flows, not sitting directly in the critical path.

Ready to Get Started?

Let's build your Customer Support Agent using the principles of Architecting for Agents! You're not just building an automation—you're building a digital colleague that can reason, adapt, and handle complex, multi-faceted customer problems autonomously.

Click Next below to begin with the prerequisites.