AI Release Notes Agent

1. One key problem which I have seen in Tinuiti Project is that when marketing platforms like Google Ads, Microsoft Ads, Reddit Ads etc. rename, remove or modify fields, it can silently break data pipelines, dashboards and data warehousing systems. These changes often go unnoticed until reporting errors occur, which impacts business decisions. 2. Manual tracking of release notes is time-consuming. 3. Critical updates often go unnoticed. 4. Field renames/deletions break data pipelines & dashboards and impacts data warehousing.

To solve problems mentioned in Problem Statement, I built an AI-powered Release Notes Agent. It’s like a bot that reads a marketing platform Release Notes webpage and picks only the latest and important updates. This is designed to help marketing teams stay up to date with frequent platform changes across Google Ads, LinkedIn Ads, Microsoft Ads etc. Instead of manually reviewing lengthy and technical release notes, the agent automatically collects updates and uses AI to generate clear, actionable summaries. Each update is categorized by impact, risk level, required actions and the teams that should pay attention, enabling faster and more informed decision-making. The solution is built using Python, integrates with an LLM for AI-driven summarization, uses Google Sheets APIs for storage and Streamlit for building the interactive dashboard interface. For generating AI insights, I have used Meta’s Llama 3 instruct model via Cloudflare AI. We have two versions of this solution: 1. A Google Sheets-based version that automatically logs updates, categorizes risk and provides structured AI summaries for teams that prefer a lightweight and shareable format. The system runs entirely in the cloud using GitHub Actions, ensuring zero manual effort. It also sends an automated email with a clean summary, so stakeholders don’t have to check multiple sources manually. 2. A UI-based dashboard that presents the same insights in a more interactive and visual way - highlighting impact, required actions, risk levels and the teams that should care, all in one place. Sheets works well for operational tracking and collaboration, while the UI is better for quick insights and executive visibility. Together, these formats ensure both operational teams and leadership can access relevant information efficiently, reducing effort, minimizing risk and improving responsiveness to platform changes.