Amazon Connect is a cloud-based contact center service that provides voice and chat capabilities with integrated AI. It handles the full contact center stack: telephony, IVR (Interactive Voice Response), queue management, agent routing, real-time and historical analytics, and workforce management. A single Connect instance can handle up to 10,000 concurrent active voice contacts by default (adjustable via quota increase). For AI projects, Connect is the deployment platform for conversational AI in voice and chat channels.

Official documentation: https://docs.aws.amazon.com/connect/
Pricing: https://aws.amazon.com/connect/pricing/
Service quotas: https://docs.aws.amazon.com/connect/latest/adminguide/amazon-connect-service-limits.html

Core Concepts

Instance - A Connect deployment with its own phone numbers, agent accounts, queues, and configuration. An instance can handle thousands of concurrent calls and chats.

Contact Flow - A visual workflow that defines how a customer interaction is handled. Contact flows control everything from the initial greeting through IVR menus, data lookups, queue placement, and agent connection. Flows are built in a drag-and-drop editor with blocks for playing prompts, getting customer input, invoking Lambda functions, and transferring to queues or agents.

Queue - A holding area for contacts waiting for an agent. Queues have routing profiles that determine which agents can handle contacts from that queue, and priority settings that control the order in which contacts are answered.

Agent - A human handler. Agents log into the Connect Contact Control Panel (CCP) to receive calls and chats. The CCP can be embedded in CRM applications using the Connect Streams API.

AI Integration Points

Amazon Lex - Embeds conversational bots directly into contact flows. When a customer calls, a Lex bot can handle the entire interaction for simple requests (checking order status, making payments, scheduling appointments) or collect information before routing to a human agent with context. This reduces average handle time and deflects calls from the agent queue.

Amazon Bedrock - Connect integrates with Bedrock for generative AI capabilities. During a live interaction, Bedrock can generate response suggestions for agents based on the conversation context and knowledge base content. Post-interaction, Bedrock summarizes the conversation automatically, eliminating manual after-call work.

Contact Lens - An AI-powered analytics feature built into Connect. Contact Lens provides real-time transcription (AWS does not publish a single accuracy figure — accuracy depends on domain vocabulary, audio quality, and accent; enterprise deployments typically achieve 85–95% word error rate reduction compared to no transcription), sentiment analysis, issue detection, and automated quality scoring. Supervisors can monitor live calls with real-time alerts when sentiment drops or specific keywords are detected. Post-call analytics include full searchable transcripts, sentiment trends, and talk-time ratios. Contact Lens real-time alerts deliver within approximately 5 seconds of the triggering event.

Amazon Q in Connect - An AI assistant for agents that surfaces relevant knowledge base articles, recommended responses, and step-by-step guides in real time during customer interactions. It uses the conversation context to proactively suggest answers, reducing the need for agents to search through documentation manually.

Common Automation Patterns

Self-service with escalation - Lex handles tier-1 queries. When the bot cannot resolve the issue, it transfers to an agent queue with all collected information (intent, slot values, conversation history) attached to the contact. The agent sees the full context without asking the customer to repeat information.

Intelligent routing - Lambda functions in the contact flow look up customer data (CRM, order system) and route based on context: premium customers go to a specialized queue, technical issues go to technical support, and billing questions go to billing. This reduces transfers and improves first-contact resolution.

Post-contact automation - Contact Lens generates a transcript and summary. A Lambda function processes the summary to create tickets in the issue tracking system, update CRM records, and trigger follow-up workflows based on detected issues.

Analytics and Reporting

Connect provides built-in real-time and historical analytics dashboards. Key metrics include service level (percentage of contacts answered within a target time), average handle time, abandonment rate, and agent occupancy. These metrics can be exported to S3 for custom analytics in Athena or QuickSight.

Contact Lens adds AI-driven analytics: theme detection (identifying trending issues across contacts), sentiment trends, and conversational analytics that go beyond traditional contact center metrics.

Pricing

Connect charges per minute of use (voice) or per message (chat), plus per-minute charges for telephony. Contact Lens has additional per-minute charges for real-time and post-call analytics. There are no upfront costs, seat licenses, or minimum commitments. This pay-per-use model makes Connect particularly attractive compared to traditional contact center platforms that require multi-year commitments.