How to Build Your Own Amazon Lex

Leverage large language models to resolve slot values when native NLU is unable to, handling free-form responses

Convert speech input to text with high accuracy, trained on telephony audio

Enable conversational answers to commonly asked customer questions using retrieval augmented generation (RAG) with foundation models

Create voice and text interfaces powered by generative AI with advanced natural language capabilities

Advanced NLU technology to understand user intent from natural language inputs

Allow users to interrupt and override bot responses for more natural conversational flow

Native context management for multi-turn conversations without requiring custom code

Intelligent pause management for more natural sounding conversations

Build sophisticated multi-turn conversations where users are prompted for required information across multiple exchanges

Enable natural conversations with pauses and interruptions using streaming conversation APIs

Leverage foundation models from Amazon and leading AI startups through Bedrock for generative AI capabilities

Native integration with AWS omnichannel cloud contact center for handling customer queries over chat or phone

Integrate intelligent search powered by machine learning to find accurate answers from unstructured documents and FAQs

Convert text responses to lifelike speech using standard or Neural Text-to-Speech voices

Native integration with AWS Lambda for data retrieval, updates, and business logic execution

Analyze existing conversation transcripts to propose initial bot design with intents and slot types automatically

Create a full bot based on a natural language description provided by the user, generating a baseline bot that can be refined

Drag-and-drop conversation builder that accelerates bot building with no-code environment and visual feedback

Pay only for what you use with no upfront commitments or minimum fees, metered per API request

Billing model where each user input (voice or text) is processed as a separate API call

Alternative billing model where all user inputs across multiple turns are processed in one streaming API call

Create multiple aliases for bots and associate different versions such as production, development, and test

Apply versioning to Intents, Slot Types, and Bots with rollback mechanisms for multi-developer environments

Speech recognition engine trained on telephony audio for increased accuracy in telephone-based interactions

Easily publish bots to chat services directly from the Amazon Lex console