Icon source: AWS

CLOUD SERVICE · AWS

Amazon Augmented AI (Amazon A2I)

Amazon Augmented AI (Amazon A2I) is a service designed to make it easier to build the workflows necessary for human review of machine learning predictions, allowing developers to integrate human judgment into their AI-driven applications.

Cloud Services Hub →

What is Amazon Augmented AI (Amazon A2I)

Read the extensive description

Amazon Augmented AI, frequently abbreviated as Amazon A2I, is an innovative service designed by Amazon Web Services (AWS) to make it easier for developers and businesses to build applications that blend human judgment with artificial intelligence (AI). This blending is particularly crucial in scenarios where AI alone cannot make decisions with high confidence.

By incorporating human review, Amazon A2I aims to improve the reliability and accuracy of machine learning (ML) model predictions, especially in complex decision-making tasks that require nuanced understanding or subjective evaluation. Amazon A2I operates by providing a straightforward pathway for developers to create workflows that determine when a human review is necessary. Upon reaching a specific threshold of confidence level in the AI's decision-making threshold that is customizable by developers the system automatically routes the task to human reviewers. These reviewers could either be in-house experts or external contractors from Amazon Mechanical Turk, a crowdsourcing marketplace that tasks humans with jobs computers are currently unable to do effectively. One of the standout features of Amazon A2I is its flexibility and integration capabilities.

It is designed to work seamlessly with other AWS services, such as Amazon Rekognition for image and video analysis, Amazon Textract for extracting text and data from scanned documents, and Amazon Comprehend for natural language processing. This extensive interoperability allows organizations to implement human reviews in various use cases, ranging from content moderation and text extraction to sentiment analysis and medical diagnosis validation.

The human review interface is also a notable aspect of Amazon A2I. It is customizable, enabling organizations to design or select pre-built worker task templates that are tailored to specific tasks. This ensures that reviewers have all the necessary context and guidelines to make informed decisions, thereby improving the effectiveness and efficiency of the review process.

In essence, Amazon A2I represents a significant step forward in the practical application of AI and ML by acknowledging and addressing their current limitations. Its ability to combine human intuition and intelligence with the scalability and speed of machine learning models brings a new level of accuracy and trustworthiness to automated decision-making processes. By doing so, Amazon A2I not only enhances the performance of AI applications but also extends their applicability across a wider range of industries and scenarios where precision is paramount.

Key Amazon Augmented AI (Amazon A2I) Features

Amazon Augmented AI (Amazon A2I) facilitates seamless human review workflows for machine learning predictions, allows easy integration with external workforces or Amazon Mechanical Turk, supports custom workflow creation for a wide range of applications, offers real-time human review capabilities, and ensures enhanced data privacy and security.

Human Review Workflow

Amazon A2I provides built-in human review workflows for Amazon Rekognition and Amazon Textract, allowing developers to add human reviews seamlessly for machine learning predictions to ensure data accuracy and quality.

Easy Integration with External Workforces

Facilitates the integration with Amazon Mechanical Turk or private workforces, offering flexibility in choosing the right reviewers for specific tasks based on skill level or other criteria.

Custom Workflow Creation

Enables the creation of custom workflows for machine learning applications that are not natively supported, allowing for a broad application of human review across many machine learning models and tasks.

Real-Time Human Review

Provides real-time human review capabilities, enabling immediate human intervention when a machine learning model's confidence score falls below a certain threshold.

Enhanced Data Privacy and Security

Implements robust security measures to protect sensitive data throughout the human review process, ensuring data privacy and compliance with industry standards.

Amazon Augmented AI (Amazon A2I) Use Cases

Amazon A2I facilitates the efficient processing of medical records, content moderation on social media, financial document handling, and customer support ticket triage, by combining machine learning with human review for higher accuracy and compliance.

Medical Record Analysis

Amazon A2I enables healthcare professionals to efficiently analyze and extract relevant data from patient records. By integrating machine learning models for initial analysis and using human oversight for validation and correction of the outputs, the accuracy of patient data processing is significantly enhanced, improving patient care and streamlining administrative processes.

Content Moderation for Social Media

This use case involves the implementation of Amazon A2I to review and moderate user-generated content on social media platforms. It combines automated flagging of potentially inappropriate content with human reviewers who make the final decision, ensuring compliance with platform policies while maintaining a high throughput.

Financial Document Processing

Amazon A2I is employed to streamline the extraction and processing of data from various financial documents, such as invoices, receipts, and contracts. Automated processes handle the bulk of data extraction, while human reviewers ensure accuracy, compliance, and handle exceptions, significantly reducing processing time and errors.

Customer Support Ticket Triage

In this use case, Amazon A2I is used to classify and route customer support inquiries. Automated algorithms perform initial classification based on inquiry content and urgency, then human reviewers adjust classifications as needed. This optimizes response times and ensures customers receive the appropriate level of support.

Amazon Augmented AI (Amazon A2I) pricing models

Amazon A2I pricing consists of charges for each human review task and additional fees for data processing and storage.

Data Processing and Storage Pricing

Charges apply for processing and storing data within Amazon A2I. This includes fees for any temporary storage of data used during the human review process, as well as any data processing needed before or after the review.

Human Review Pricing

Costs are incurred based on the number of human review tasks created. You pay per task, which varies depending on the complexity and the time spent by human reviewers on each task.

Services Amazon Augmented AI (Amazon A2I) integrates with

Amazon Rekognition

A2I integrates with Amazon Rekognition to enable human review for image and video analysis tasks. Use cases include label verification, face comparison, and content moderation.

Amazon SageMaker

A2I integrates with Amazon SageMaker to provide automated machine learning workflows where human review is required. SageMaker provides the model predictions that can be reviewed and adjusted through A2I's human review workforce.

Amazon Textract

A2I works with Amazon Textract to facilitate human review workflows for extracting text and data from scanned documents. Primarily used to verify complex data extractions that require human judgment.

AWS Lambda

A2I integrates with AWS Lambda to trigger custom functions in response to events within the human review process. This can be used to automate actions based on the human review results.

Amazon S3

Amazon S3 is used to store datasets and result outputs for A2I human review workflows. This integration simplifies the data transfer and storage in a secure and scalable manner.