SentimentAI Dashboard
Real-time AI-powered social media sentiment analysis platform — GPT-4o classifies 500+ posts/day per brand, cutting client reporting from 2 days to 4 hours.
Client background
BrandPulse is a social media analytics and reputation management agency based in Dubai, serving 8 enterprise clients in the F&B, hospitality, and retail sectors. Their differentiation was depth of insight — but the manual analysis process was crushing their margins.
The challenge
Four analysts were manually reading, categorising, and summarising social media mentions for client brands — across Twitter/X, Facebook, Instagram, Reddit, and Google Reviews. Each analyst handled 2 clients, spending 2 full days to produce a weekly report. With new clients wanting real-time alerts, the manual process had become a ceiling on growth.
Discovery & planning
Operations Review
Observed an analyst producing a real report for one client. Documented every step: collection, reading, categorisation, Excel tagging, writing summary. Measured: 9 hours per client per week.
Accuracy Benchmarking
Analysts rated 200 sample posts for sentiment (positive/negative/neutral/mixed). We tested GPT-4o against the same 200 posts. GPT-4o reached 93% agreement with analyst consensus.
Data Source Planning
Mapped available APIs for each platform. Twitter API v2, Facebook Graph API, Reddit API, and Google My Business API all required different auth flows. Planned scraper fallbacks where APIs were restricted.
Dashboard UX Design
2-hour wireframing session with BrandPulse's head of client services. Designed the real-time dashboard, weekly digest email, and critical alert notification format.
Technical solution
We built an always-on data pipeline that collects social mentions for each monitored brand every 30 minutes, passes them through a LangChain chain that classifies sentiment (5-point scale), extracts themes, and identifies key quotes, then stores structured results in PostgreSQL. A React dashboard shows live sentiment trends, topic clusters, and competitor comparisons. Analysts now review AI output and add strategic commentary — the research is done.
Tech stack
Timeline
Twitter/Facebook/Reddit API integration, Playwright fallback scrapers, data normalisation schema
LangChain chain design, GPT-4o prompt engineering, accuracy benchmarking (95% target achieved), structured output parsing
React dashboard, sentiment charts, topic clustering, PDF report generator, Slack alert integration
End-to-end testing with all 8 client brands, analyst training, documentation, production deployment
Results & outcomes
Our analysts now focus entirely on strategy and client relationships — the AI handles the reading, classifying, and summarising. The accuracy genuinely surprised us. Two clients asked us to pitch our competitors because of this capability.
Book a free call and we'll scope your project in 30 minutes.
Book a free call