Wendy Wong for AWS Heroes

Posted on Dec 30

Build AI agents and automate UI workflows - Amazon Nova Act and Kiro

#agents #tutorial #ai #aws

Reimagine the possible with AI agents

Building workflows takes time and effort to maintain for developers and engineers. With designing automated workflows we need to consider our business processes and logic. At AWS re:Invent 2025, on Day 3, AWS VP of Agentic AI Dr Swami Sivasubramanian announced the general availability of Amazon Nova Act in the AWS region United States (N.Virginia) us-east-1.

Lesson Objectives

In this lesson, you will learn how to:

Build an AI agent from UI automated workflows
Identify use cases for Amazon Nova Act

What is Amazon Nova?

AWS introduces a family of AI offerings under Amazon Nova which include the following:

a) Amazon Nova 2 Foundation Models:

Nova 2 Lite - This is foundation model cost effective for reasoning tasks such as building chatbots. The data type includes: text, video and image.
Nova 2 Pro (preview) - This is an intelligent reasoning model suitable for AI tasks such as agentic coding. The data type includes: text, video,image and speech.
Nova 2 Omni (preview) - This foundation model is capable of
Nova 2 Sonic - This is a speech-to-speech model that can be used for conversational AI. The data type includes: text and speech.
Nova multimodal embeddings - This is a multimodal embedding model that can be used for agentic RAG and semantic search. The data type includes: text, image, audio and video.

b) Frontier Model:

Amazon Nova Forge - This frontier model allows you to combine your organizations' data with curated training data from Amazon Nova.

c) Automate web browser workflows at scale

Amazon Nova Act - This AWS service that built on the architecture of Nova 2 Lite to create AI agents to automate web browser UI workflows.

What is Amazon Nova Act?

Amazon Nova Act is an AWS service that helps you to build and manage AI agents tasked with automating UI workflows at massive scale for repetitive tasks.

Take a closer look at how Amazon Nova Act is used at Amazon AGI Lab and how it become generally available for web browser automation to assist with taking action and making decisions.

What are the features?

With web browser automated workflows the capabilities of Amazon Nova Act include:

Extracting information
Confirming bookings
Filling out forms
Perform a web search

Agents may be quickly created with natural language prompts and python code.

What are the benefits?

Amazon Nova act completes repetitive tasks, launches APIs and can also perform escalations.
Develop UI automation agents from prototype to production in a matter of hours.
Deploy and manage fleets of agents

What are the use cases?

The critical UI workflow automation from your browser can solve tasks such as:

Data entry - Use agents to help you populate details and remove the need to copy and paste data.
Web QA testing - validating user journey workflow in a web browser instead of using test scripts.
Checkout flows - the agent may take actions to complete a workflow for example a purchase, refund, validation and UI changes.
Data extraction - navigate websites and extract data from unstructured data sources that cannot be programmed or easily exported.

Task: Build an AI agent to assist a first-year psychology student enrol into an Australian university

Tutorial 1: Build reliable AI agents for UI workflow automation with Amazon Nova Act and Kiro

Step 1: Launch Kiro (i.e. agentic IDE) from your desktop and sign in with your AWS Builders ID.

Allow Kiro to access your data. Select Allow access.

Step 2: In Kiro, navigate on the left-handside to Extension and search for Amazon Nova Act to install the extension.

Select Trust publisher and install to install the extension.

Step 3: Now start developing a prototype to build an AI agent with Amazon Nova Act for automating UI workflow from a web browser.

You may experiment with your user journeys in the Amazon Nova Act Playground and generate an API key from the link:

https://nova.amazon.com/act

I am not able to access this playground from my location in Sydney (ap-southeast-2)

Step 4: Create your first workflow

With the Nova Act extension installed, in the Kiro terminal enter:

Upgrade to the latest version

pip install --upgrade nova-act

Step 5: Create a project folder a name it 'UTS enrolment folder' and save it in your working directory such as Documents

Step 6: In Kiro IDE go to File -> Open Folder -> UTS enrolment folder

In Kiro select vibe coding and develop a prototype in this agentic IDE and enter this prompt:

Kiro will create a prototype and save the files in your project folder.

You can open the 'uts-psychology-enrollment-workflow' html file and have a look at the prototype and share with your team.

You may also inspect the README file.

Step 4: You may create the Steering files using Kiro.

Step 5: You may also create your SPECS using Kiro.

With a staged process, Kiro will design the:

*a) Requirements *

b) Design document

c) Analyze the acceptance criteria

d) Design Implementation Plan

If you inspect your project folder, you will be able to see a subfolder titled '.kiro' automated and it includes both the steering files and the specs.

Kiro has completed all the specifications for the development from the beginning.

Tutorial 2: Deploy the web browser workflow with Amazon Q in Visual Studio

Step 1: Open Visual Studio Code and Open a Folder 'UTS enrolment folder'.
Step 2: Open chat in Visual Studio and enter the prompt.

Step 3: In Visual Studio with Amazon Q we can deploy the workflow into AWS via our IAM credentials from the specs created in Kiro.
Two deployment files were created:
** Deployment script**
A deployment.md file was created

You may also deploy this workflow on Amazon Nova Act CLI and you may productionize this workflow on Nova Act CDK templates.

More learning resources

Until the next update, happy learning! 😀

DEV Community