Amazon Nova Act, introduced by Amazon AGI Labs, is a cutting-edge AI model for automating tasks within web browsers. It promises to enhance productivity by effortlessly handling activities, from setting appointments to managing out-of-office emails, making it a valuable tool for efficient workflows.
Key Points:
- Automates bookings, orders, and calendar holds.
- Interacts with varied web UI elements, including games.
- Consistently high performance in internal benchmarks.
- SDK available for developers to build reliable task agents.
- Planned integration with Alexa+ for autonomous internet navigation.
Amazon Nova Act: AI for Automating Web Tasks
Amazon Nova Act is a groundbreaking AI model introduced by Amazon AGI Labs. It’s designed to perform actions within web browsers, revolutionizing task automation. Released as a research preview on March 31, 2025, Nova Act aims to serve as an AI agent for web automation tasks.
What Can Amazon Nova Act Do? Key Features and Use Cases
Amazon Nova Act’s core functionality lies in its web browser automation capabilities. This model can handle a wide array of tasks, including:
- Booking reservations
- Ordering food
- Submitting out-of-office requests
- Placing calendar holds
- Setting up ‘away from office’ emails
Amazon Nova Act excels at understanding and interacting with diverse UI elements across different environments. For example, it can engage with web games even without prior specific training, showcasing its versatility.
So, what is Amazon Nova Act? In essence, it’s an AI agent capable of performing practical, everyday tasks within web browsers, making it indispensable for users seeking more efficient workflows.
Amazon Nova Act Performance: Benchmarks Against Competitors
Amazon’s internal evaluations highlight Nova Act’s superior performance in web automation tasks. For example, it scored an impressive 94% on screen interaction benchmarks, demonstrating reliability in tasks like date picking, drop-down selections, and handling pop-ups.
Benchmark Comparison Table
| Benchmark | Amazon Nova Act | Claude 3.7 Sonnet | OpenAI CUA |
|---|---|---|---|
| ScreenSpot Web Text (Follow natural language instructions to interact with a textual element on screen, e.g., set font size to 50) | 0.939 | 0.900 | 0.883 |
| ScreenSpot Web Icon (Follow natural language instructions to interact with a visual element on screen, e.g., how many stars does this GitHub repo have?) | 0.879 | 0.854 | 0.806 |
| GroundUI Web (Understand and interact with various UI elements on the web) | 0.805 | 0.825 | 0.823 |
All benchmarks were measured internally by Amazon.
How Does Amazon Nova Act Work? Tools for Developers via SDK
The Amazon Nova Act SDK for developers is available at nova.amazon.com, offering experimentation tools. From a developer’s perspective, how does Amazon Nova Act work? The SDK provides the following:
- Tools to build agents capable of completing tasks in a web browser
- Utilizes atomic command structures to break complex workflows into reliable commands
- Allows adding detailed instructions to commands for enhanced reliability
- Supports interleaving Python code for tests, breakpoints, asserts, or parallelization
This structure addresses limitations often faced with web page load times, making it highly reliable for developers.
Powering the Future: Nova Act and Alexa+ Integration
Amazon has ambitious plans for Amazon Nova Act Alexa+ integration. Nova Act will power features in the upcoming Alexa+ upgrade, significantly enhancing Alexa+‘s capabilities. This integration will enable Alexa+ to navigate the internet autonomously to complete tasks, especially when integrated services lack necessary APIs.
An example use case includes automating scheduled food delivery orders, demonstrating its potential to streamline daily tasks further.
Industry Significance and Community Feedback on Nova Act
Nova Act represents a significant advancement in AI that can handle complex, multi-step autonomous tasks. This development strengthens Amazon’s competitive edge in the AI assistant market, pitting it against competitors like OpenAI and Anthropic.
External feedback has been positive. Mindplex Magazine noted Nova Act’s potential to challenge existing AI assistants, highlighting its reliable web automation features as a key strength. Nova Act stands out as a powerful tool for developers, marking an important step forward for AI-driven web automation.
