Support that carries the context

What makes this difficult

• Support outlasts staffed hours

  Customers can need routine information when agents are unavailable, but a small business cannot assume human coverage at every hour.

• Questions vary in wording

  Customers ask similar questions in different ways, making rigid decision trees awkward when the answer already exists in business help content.

• Handoffs can lose the issue

  An unresolved chat should reach a person with its context intact, rather than making the customer explain the same problem again.

The hard part

A handoff should continue the same issue

The difficult boundary is between an AI answer and a person taking responsibility for the unresolved issue. Connect keeps the interaction context as assistance moves to an agent, while Q in Connect grounds self-service in the business knowledge base. This supports continuity, but useful escalation still depends on current help content and an available human support process.

Why this design

1. Ground conversational assistance

   Chosen

   Use Amazon Q in Connect with help content stored in S3.

   Because

   Customers need answers to naturally phrased questions about the business policies and troubleshooting guidance.

   Cost

   Generative responses require maintained knowledge content and quality review.

   Reconsider when

   Use a tightly scripted interaction when the task requires a fixed sequence with strictly controlled wording.

2. Embed the support entry point

   Chosen

   Place the Amazon Connect communication widget in the business website.

   Because

   Customers can ask for help from the page where the question arises.

   Cost

   The website experience and support configuration must be maintained together.

   Reconsider when

   Use a separate portal when support needs a broader authenticated workspace beyond a conversation.

3. Combine self-service and agents

   Chosen

   Escalate unresolved conversations to human assistance with their context.

   Because

   Routine questions and exceptional cases need different kinds of support.

   Cost

   The business still needs staffing, routing and ownership for escalated issues.

   Reconsider when

   Simplify the channel mix when the team cannot reliably support all offered escalation options.

Cost and operating assumptions

• Customer contacts

  Connect usage depends on the channels used and their applicable billing units, including voice minutes and chat messages or sessions.

• AI assistance

  Amazon Q in Connect adds usage-based charges under the selected service options.

• Website and knowledge storage

  S3 stores website assets and help content, with storage and request activity contributing to cost.

• Website delivery

  CloudFront delivery charges depend on website traffic and the applicable delivery options.

How it stays reliable

• Keep the conversation context

  The guidance carries context into agent escalation across supported channels, reducing the need for customers to restart their explanation.

• Retain the service record

  Connect captures interaction history for review and service improvement across AI and human assistance.

• Maintain the knowledge source

  Design consideration: review the S3 knowledge content as policies change, because grounded responses can still reflect outdated source material.

• Plan human availability

  Design consideration: around-the-clock self-service does not establish around-the-clock agent coverage, so escalation expectations must match staffing.

When this is a poor fit. A poor fit when the business has no maintained help content or no reliable human process for handling questions that self-service cannot resolve.