Agentic AI for Connected Support

• 

  Title

• 

  Overview

• 

  What we Built

• 

  Troubleshooting Journey

• 

  Capability Delivered

• 

  Solution

• 

  Under the hood

• 

  contact-center-flow-account-page

• 

  contact-center-flow-account-page2

• 

  Auto Diagnostic, case discovering and account snapshot

• 

  alternative solution if tutorial did not work

• 

  Showing top 3 solution

• 

  Guided Screen Flow launched from Agentforce

• 

  Tutorial Launch from Agentforce

• 

  Case close

• 

  Conversation Summarization

• 

  Self Care Demo 1

• 

  Self Care Demo 2

• 

  call transfer from self care to contact center - 1

• 

  call transfer from self care to contact center - 2

• 

  call transfer from self care to contact center - 3

• 

  call transfer from self care to contact center - 4

• 

  Telegram Demo

💡 Inspiration

In today’s telecom landscape, support agents often juggle multiple disconnected tools while customers experience inconsistent troubleshooting paths. These inefficiencies result in long resolution times and poor customer satisfaction.

Inspired by this challenge, we envisioned an AI-powered assistant that works across channels — guiding both agents and customers through intelligent, real-time resolutions using Salesforce Agentforce. Our goal was to deliver consistent, context-aware support that’s both scalable and frictionless.

🤖 What it does

Our solution, CATRE (Customer and Agent Troubleshooting Resolution Experience), is an Agentforce-powered assistant that:

• Automatically runs diagnostics based on the customer’s BAN or device state
  
• Summarizes open tickets and system findings
  
• Recommends the top 3 troubleshooting paths using a GenAI intent classifier
  
• Displays diagnostic summaries and top solution paths in rich text for intuitive user experience
• Delivers step-by-step resolution guidance from knowledge articles
  
• Supports seamless session handoff between self-care and contact center
  
• Works across three support channels:
  
  • Contact Center Agent (Agentforce for Employees)
    
  • Customer Self-Care (Experience Cloud site)
    
  • Messaging Channel (Telegram Bot)

🛠️ How we built it

We built CATRE using Salesforce-native tools with a flexible agentic design:

• Agentforce + Prompt Templates to classify intents, summarize diagnostics, and provide AI-guided responses
  
• Custom Apex & Flows to launch guided flows, reset services, manage cases, and log interaction summaries
  
• Optimal Journey Object to track customer troubleshooting sessions and enable real-time handoff to agents
  
• Data Cloud to fetch and enrich account-level diagnostics
  
• Platform Events to launch tutorials or Salesforce flows in tabs
  
• Telegram Integration via webhook + Apex connector to invoke Agentforce natively
  
• Lightning Components embedded for internal and Experience Cloud UIs
  

🧗‍♂️ Challenges we ran into

• Maintaining session continuity across web, contact center, and messaging
  
• Handling callouts and DML limits in guest user flows
  
• Dynamically switching instructions based on agent vs. customer roles
  
• Managing statefulness and memory in messaging integrations like Telegram
  
• Ensuring prompt reusability without leaking backend or diagnostic jargon
  

🏆 Accomplishments that we're proud of

• Built and deployed an AI assistant across three distinct channels
  
• Delivered a true multi-modal GenAI agent using only Salesforce-native tools
  
• Designed a session logging model (Optimal Journey) for real-time context sharing
  
• Created smart summaries and decision points using custom Agent Actions
  
• Enabled real-time escalation to guided flows, support case creation, or live agents
  

📚 What we learned

• A well-designed prompt can reduce friction more than 100 lines of logic
  
• Agentforce’s low-code/no-code integration is incredibly powerful when orchestrated well
  
• Even GenAI needs structure — memory management and escalation logic are essential
  
• End-user experience (especially in self-care) requires simplification, not just replication
  
• Omni-channel support can be delivered with strong session design and event-based communication
  

🚀 What's next for Agentic AI for Connected Support

• Expand to Slack, WhatsApp, or VoiceBot channels
  
• Integrate with product catalog and CPQ for guided upsell/plan-change conversations
  
• Add translation and accessibility enhancements
  
• Support other telecom services like Broadband, IPTV, IoT
  

Built With

• apex
• data-cloud
• experience-cloud
• flex-ui
• html
• lightning-flows
• lwc
• platform-cache
• platform-events
• prompt-templates
• rest-api
• salesforce-agentforce
• salesforce-custom-objects
• telegram-bot-api
• visual-studio-code