Case Study: UpdateAI achieves 180% improved action-item detection and real-time transcription speed with Deepgram

Eliminating busywork and building customer trust in online meetings

UpdateAI, a provider of AI meeting assistants for Customer Success teams, faced a growing burden of administrative work from back-to-back Zoom meetings and needed to capture action items in real time. They required an ASR that offered real-time speed, >90% accuracy, and scalable cost — after testing six providers they selected Deepgram’s AI Speech Recognition for its superior accuracy and transcription speed.

Using Deepgram’s speech recognition as the data foundation, UpdateAI built an NLP action-item engine that pattern-matches positive signals and filters out false positives; the combined solution boosted action-item detection accuracy by over 180% versus competitors in tests of 25 meetings (1,000 minutes), delivered 120x faster pre-recorded transcription (1 hour → 30 seconds), and produced more accurate Zoom notes—freeing CS teams to focus on relationships and upsell opportunities. UpdateAI is continuing to work with Deepgram on tailored speech models to further improve accuracy for accents, dialects, and domain terms.

UpdateAI

Josh Schachter

Chief Executive Officer