The Stack Overflow Podcast How to create agents that people actually want to use
6 snips
Nov 18, 2025 Assaf Elovic, Head of AI at monday.com, shares insights on building AI tools people genuinely want to use. He discusses the initial failures of their chatbot and the importance of aligning user expectations with product design. Elovic emphasizes that user adoption outweighs technical perfection and explains how they revamped Monday Sidekick for better user experience. He highlights the potential of autonomous agents and voice technology, discussing how these innovations can enhance productivity while ensuring trust in automated systems.
AI Snips
Chapters
Transcript
Episode notes
Early Chatbot Misfire
- monday.com launched an open-ended chatbot called Monday Expert and found production accuracy far lower than lab tests.
- The bot guided users to knowledge-base articles 80% of the time, turning it into a frustrating support tool and causing high churn.
Plan For The 99% You Can't Predict
- Open-ended conversational products must plan for the vast majority of inputs they won't handle.
- Defining limited 'correct' answers ahead creates bias and misses real user behavior.
Trigger Agents In Context
- Narrow context reduces user expectations and improves success rates.
- Place the assistant entry point inside a specific task or board item instead of the global nav.