Jonathan Weizman

📈 Hotel guest feedback centralized with Suite Talk using n8n, integrates Google and TripAdvisor reviews for efficient reporting. Hourly workflow automates sentiment analysis via OpenAI, flags negative experiences, and creates Zendesk tickets if needed. It standardizes data across platforms, saves information to Postgres, and suits hotel groups and CX teams. Accessible on GitHub, it's configurable for fast triage of guest feedback.

📈 Built a system to organize hotel reservation notes, eliminating misplaced details. Key steps: 1) Developed an extraction layer, 2) Identified housekeeping and maintenance tasks, 3) Automated task workflows, 4) Ensured no duplication or oversight. Result: saved hours of daily work. Compatible with Mews, Apaleo, or Cloudbeds. Interested parties can implement this setup in their hotels.

🧠 RAG (Retrieval Augmented Generation) enhances AI response accuracy by grounding it in specific business data. Traditional computers fail to understand nuanced language, while AI translates semantic proximity into spatial vectors, even in thousands of dimensions. Large Language Models (LLMs) need guidance to avoid inaccuracies. By feeding LLMs with actual documents like HR policies or hotel information, RAG creates a personalized AI knowledge base, improving customer service and maintaining natural language understanding. The next episode will discuss Transformers in AI. The article also mentions the use of Perplexity.ai as a superior alternative to traditional search engines for complex queries.

💻 Jonathan Weizman, Hotel Tech CEO at Roomchecking.com and a developer, discusses AI's impact on the hotel industry and his daily focus on utilizing it. He highlights the shift from traditional Natural Language Processing (NLP) to advanced AI tools like ChatGPT and Large Language Models (LLMs), which reason through prediction rather than set responses. These models are trained on trillions of words, digesting diverse internet content to learn and predict. The article touches on the phenomenon of AI 'hallucination', where AI predicts plausible text regardless of its truthfulness, exemplified by ChatGPT's over-representation of the number "42" due to its cultural significance.