Skip to content

Lab2.md

Latest commit

 

History

History
126 lines (87 loc) · 4.21 KB

Lab2.md

File metadata and controls

126 lines (87 loc) · 4.21 KB

Building Up Call Center Using Amazon Connect & Lambda in Node.js Environment with JavaScript

☎️ Lab 2 – Building a Call Center with Amazon Connect

🧭 Goal

Deploy a basic cloud-based call center using Amazon Connect, hook it up with our PersonalBanker Lex bot, and route calls either to the bot or a live agent. Sprinkle in a few enhancements like DTMF options (press 1 / press 2) and personalized greetings.

📞 Setting Up Amazon Connect

I launched Amazon Connect in North Virginia (same region as Lex).

Key setup points:

  • Skipped admin creation (used default)
  • Disabled outbound calling
  • Took the default S3 bucket settings
  • Created the instance and claimed a UK Direct Dial Number (for this lab)
  • Noted down the number for testing later

🌀 Creating a Simple Contact Flow

Instead of editing the huge sample flow, I imported a lightweight version:

  • Downloaded Simple Workflow file from the repo
  • Opened Routing → Contact Flows
  • Clicked Create contact flow, then used the Import Flow option
  • Loaded the Simple Workflow file
  • This flow: greets the caller → plays a prompt → transfers them to a queue

I published it with Save & Publish, then:

  • Went to Routing → Phone Numbers
  • Edited my claimed number to link it to this new Simple Workflow

Now when I called in, it said the welcome message and sent me to a queue. Neat, but we’re just getting started.

🤖 Integrating the Lex Bot

Now came the fun part—injecting my PersonalBanker Lex bot from Lab 1 into this call flow.

Steps:

  1. Back in Amazon Connect Admin, I added the Lex bot under Contact Flows → Amazon Lex → + Add Lex Bot
  2. Then I edited the Simple Workflow flow:
    • From the left panel, dragged Get Customer Input into the flow (from the Interact category)
    • Added a prompt: 
      How can we help you today?
    • Selected Amazon Lex, picked PersonalBanker
    • Added intent: GetBalanceCheck
  3. Unlinked the Play Prompt → Queue connection
  4. Linked Play Prompt → Get Customer Input → then its Default branch → Transfer to Queue
    (Other branches like "Error" or "No Input" went to Disconnect)

Saved and Published again.

Tested it: when I called the number, I got prompted by Lex. Said “check my bank balance” → got routed into the bot → entered my account type and PIN → got back a mock balance. Mission accomplished.

🎛️ Enhancing the Experience with DTMF + Additional Intents

I realized there was no way to speak to an agent from the bot, so I:

  1. Created a new intent in Lex: Advisor
    • Sample utterances: 
      advisor  
talk  
talk to someone  
press one
  2. Added utterance press two to GetBalanceCheck as well (for symmetry)

Built the bot again in Lex.

Then back in the Get Customer Input box in Connect, I added both intents:

  • GetBalanceCheck
  • Advisor

Wired the Advisor branch to Transfer to Queue, so it now allowed users to say:

  • “talk to someone”
  • or literally press 1 on the phone pad

This added both voice and keypad navigation options. Tested again—pressing 1 got me to an agent; pressing 2 triggered the balance check flow. Felt legit!

🧑‍💻 Bonus Round: Personalized Greetings via Caller Number

I took it one step further—made Connect greet callers by name using their phone number.

Steps:

  1. Created a new Lambda: myPersonalResponder

  2. Pasted in the code from myPersonalResponder_v1.js

    • It checks the incoming number and returns a name if matched

  3. In Contact Flow editor:

    • From the left panel, dragged in Invoke AWS Lambda Function
    • Wired it right after the “Entry point” of the flow

  4. Used the returned value to set a Contact Attribute like CustomerName

  5. In the Play Prompt, used dynamic speech like:

    Hi {CustomerName}, thanks for calling!

Now the system greeted known callers by name. Otherwise, it defaulted to a generic message.

🧾 Takeaways

  • Connected Lex bots into real call flows
  • Gave users voice + keypad options for interaction
  • Handled branching logic with intent mapping
  • Personalized flows using Lambda + caller number detection