AWS Shows How Amazon Nova Act Automates Competitor Price Monitoring

Competitor price monitoring no longer has to be done manually: AWS has shown how to build an agent based on Amazon Nova Act that automatically opens retailer websites, finds the needed product, extracts prices and promotional terms, and compiles everything into a structured result. The idea is simple: instead of employees spending hours switching between tabs and spreadsheets, a company gets an almost continuous market observation loop and can make pricing decisions based on fresh data. AWS describes the typical problem faced by e-commerce teams: to understand how competitors behave, you need to regularly check dozens of product pages, manually record prices, discounts, and promotion deadlines, then transfer this data to spreadsheets.

Such a process is slow, scales poorly, and inevitably produces data entry errors. If market prices change several times a day, even a delay of a few hours turns analytics into an archive rather than a tool for making quick decisions. AWS particularly notes that similar pain exists not only for online retailers, but also for insurance companies, banks, travel and hospitality companies, where there is also a constant need to compare competitor offerings.

The key element of the solution is Amazon Nova Act, an open-source SDK for browser automation with control through natural language instructions. A developer assembles a workflow in Python from small actions: open a website, find a product, navigate to the product card, extract needed fields, verify conditions, handle errors, or add a pause. Unlike rigid scripts based on CSS selectors, this approach is designed for live websites where banners, promotional blocks, element order, and navigation constantly change.

For data extraction, AWS recommends using act_get() with a Pydantic schema so that the agent returns already validated and typed structure suitable for further loading into internal systems, dashboards, or pricing models. Particular emphasis is placed on scale. One instance of Nova Act works with one browser, but multiple instances can run in parallel.

In AWS's example, ThreadPoolExecutor is used and verification is distributed across multiple sources simultaneously. In practice, this means the agent can cover Amazon, Target, Best Buy, Costco, or any other set of websites in a single pass, then compile the overall result into one table. In the demonstration scenario, the user provides a product name and SKU, after which the agent searches for a relevant product card, distinguishes advertising results from organic ones, extracts the price, promotion details, availability, and additional metadata.

The result is recorded in CSV so it can be fed to a BI system, internal API, or dynamic pricing logic. AWS does not gloss over practical limitations. If a website displays a CAPTCHA, Nova Act does not attempt to solve it automatically.

Instead, the workflow can detect the presence of a CAPTCHA and stop so a human can complete the verification manually. For local execution, a headed mode is offered, and in a cloud scenario — human-in-the-loop through AgentCore Browser Tool with interface takeover in the AWS console. Additionally, the service returns errors as ActError, enabling retries, fallback branches, and proper logging.

For development, AWS recommends using extensions for Kiro, VS Code, and Cursor, and for monitoring runs — the Nova Act console with traces, screenshots, logs, and artifacts in Amazon S3. What this means: AWS promotes Nova Act not as a demonstration of "an agent for the sake of an agent," but as an application layer for routine web processes where speed, reproducibility, and scale matter. For retail, this is a direct path from manual monitoring to nearly continuous competitive price monitoring.

For other industries, the signal is the same: if competitor data is still being gathered through tabs and Excel, browser agents are starting to look not like an experiment, but like a practical tool.