This system provides automated analysis of news coverage for Microsoft Copilot+ devices, tracking public sentiment and specific aspects of these AI-enhanced computers across various categories. The system aggregates news from Google News, processes articles using Claude AI for detailed analysis, and generates daily summaries with quantitative scoring of different device aspects.
Motivation
I am a part of the team that launched Copilot+ devices in June of 2024. With the launch of Copilot+ devices, understanding public reception and identifying areas of strength and concern became crucial. This system was developed to:
- Track real-time public perception of Copilot+ devices
- Identify recurring positive features and potential issues
- Monitor trends in device performance, reliability, and user satisfaction
- Provide data-driven insights for product improvement
More generally, such AI-powered news analysis systems can be used for:
- Product teams tracking market reception of new technologies
- Marketing teams understanding public sentiment and messaging effectiveness
- Customer support teams anticipating potential user concerns
- Business intelligence and competitive analysis
- Research on technology adoption and public perception
- Quality assurance teams monitoring reported issues
- Product development prioritizing improvements based on user feedback
Screenshots
Link to the news analysis dashboard
(Updated daily)
Development Details
The code is created by ChatGPT and Claude. The system consists of several components:
- News aggregation module using Google News API
- Text extraction from news articles
- Multi-stage AI analysis pipeline using Claude
- Data storage and historical tracking
- Web interface for viewing results
- Automated report generation
- Trend analysis and visualization
AI Implementation
The system leverages ChatGPT 4o in three distinct stages:
- Individual Article Analysis
- Processes each article to extract positive aspects and concerns
- Uses structured table format for consistent data extraction
- Validates content relevance to Copilot+ devices
- Prompt: “Microsoft Copilot+, which is a new computer hardware platform developed by Microsoft and its OEMs (Lenovo, Samsung, HP etc.). The user will give you an article related to Copilot+. Create a table with 2 columns: 1. Key positive aspects of Copilot+ (add specifics as needed) 2. Key concerns related to Copilot+ (also add specifics as needed). Use markdown, put one item into one table cell (don’t merge items), don’t number items, don’t put a lists into a cell. Number of items in columns don’t have to be the same. Ensure the data in the table is accurate and complete. If the news is not about Copilot+ devices, just say “This is not a Copilot+ review”.”
- Daily Executive Summary Generation
- Aggregates individual article analyses
- Provides category-specific overviews
- Maintains unbiased analytical perspective
- Prompt: “You are a news analyst. The user will give you Markdown text with a summary of today’s news related to Microsoft Copilot+ (a new platform powered by AI. The platform is developed by Microsoft and OEMs). Provide analysis of these news in the form of unbiased overviews for categories like “Battery Life”, “System Performance”, “Hardware”, “Software”, “Security”, “Privacy”, “Reliability” (add others if needed). Do not use lists, but explain instead. Make category names bold in the output.“
- Scoring and Trend Analysis
- Quantifies daily sentiment across key categories
- Generates numerical scores based on mention frequency and bias
- Produces explanations for scoring decisions
- The output is used on the dashboard home page
- Prompt: “You are a news analyst for computer reviews. You analyze summaries and define scores, where one positive mention is +1, negative -1, and then you sum them up per category (if there’s nothing for a category, you say “N/A”). The below Markdown text with a summary of today’s news related to Microsoft Copilot+ (a new platform powered by AI. The platform is developed by Microsoft and OEMs). What are scores for the following categories: “Battery Life”, “System Prformance”, “Hardware”, “Software”, “Security”, “Privacy”, “Reliability” (whether the system or apps crash). Return a table formatted as a markdown with columns “Category”, “Score” and “Explanation” (with details of how/why the score was calculated).“
Learnings
<To be added>
Bottom Line
This simple system demonstrates the potential of automated AI-powered news analysis. By automatically processing and analyzing news coverage, it provides quick insights into public perception of Copilot+ devices, and could be easily extended into other areas. As a free bonus, this system creates a catalog of news, which can be used to analyze additional aspects (in my case, it was used to better understand tools and methodologies used to benchmark devices).
The system’s modular design and use of AI prompting enables adequate categorization and scoring of various device aspects, while maintaining historical context for trend analysis. In addition to internal corporate data and analytics, this became a valuable tool for our team for understanding market reception and latest reaction to news and events related to Copilot+.