How AI helps discover new competitor features before their official announcement
Marketing analyst Vladimir Mikhailov from Garage Eight created a system that detects new features in competitors' products and sends significant changes to the team.
Mikhailov replaced HTML page comparison with an analysis of what actions have become available to the user. The system opens competitors' products in an authorized browser, examines the interface, converts the discovered capabilities into a structured model, and compares it with the previous state.
This approach is intended to filter out technical noise: changes in CSS classes, URLs, DOM structure, button text, and A/B test variants. Deterministic code is responsible for comparing the prepared models, while AI is used at the stage of extracting features from the interface. As a result, monitoring records not just the fact of a frontend update, but the appearance, modification, or removal of a product's capability.
The system sends notifications about significant changes to Slack. For analytical queries, Mikhailov connected Claude via MCP: with its help, it is possible to search for new features over a selected period, compare categories and products, and check the monitoring history. The data is additionally synchronized with Google Drive and used as a knowledge base in a separate Claude Project.
To access closed features, saved Chrome profiles and Playwright MCP are used. Authorization is verified by several indicators—such as the user menu, account switcher, and logout button. The system accounts for the states of active, expired, or two-factor authentication-required sessions.
