A big upgrade for intent recognition on Voiceflow! Check out the changelog here: https://lnkd.in/gBWKwyk8
CEO @ Voiceflow | Helping Enterprise product teams build AI Agents to automate customer support and beyond
Intent classification with NLU is dead, well at least in Voiceflow 🪦 We've switched our default classification system to meaning based intent classification using RAG for a far more accurate, scalable, and multilingual classification system. Don't worry, you can still select NLU + LLM classification in the Intent settings - we've just made RAG based classification our default system. Our new intent classification system focused on meaning with less reliance on word matching so an intent can be matched more easily despite sharing 0 words of any kind. This creates far easier to maintain intents, and a more accurate classification system for better customer AI bots. Further, the new RAG classification system is infinitely scalable and can train in a second - allowing for rapid testing and iteration. Voiceflow customers can switch to the new RAG system (or to any of the 3 available classification systems in VF) with the click of a button, no migration required. ⚡ Check out what Peter Isaacs has to say in the latest Context episode 👇 #RAG #AI