October 2025 Summaries

Voice AI has transitioned from experimental to mainstream, with companies like Thoughtly and Gladia leading the charge in creating production-grade voice agents for enterprise use. In a discussion between Alex Casella from Thoughtly and Jean-Louis Quéguiner from Gladia, they explored the technical and operational intricacies of building effective voice agents, emphasizing the importance of optimizing latency, creating robust evaluation frameworks, and scaling infrastructure wisely. As the voice AI market matures, the speakers highlighted the necessity of selecting the right vendors, maintaining compliance, and fostering a culture that balances rapid iteration with technical discipline. They also discussed the diminishing advantage of self-hosting AI models due to improved performance of modern APIs, suggesting that companies should focus more on orchestration and customer experience. Both Casella and Quéguiner advised new startups to concentrate on delivering tangible value within specific verticals to achieve product-market fit and capitalize on the vast opportunities in the voice AI landscape.

As voice AI agents increasingly integrate into customer and employee interactions, ensuring their safety and accuracy is paramount to prevent reputational harm and legal risks, particularly in enterprise contexts. These AI systems often rely on large language models (LLMs), which are prone to "hallucinations," generating plausible but incorrect responses. This issue is not solely a model problem but is influenced by the architecture, including how audio is processed, knowledge is retrieved, and speech is synthesized. To mitigate risks, companies must establish robust guardrails that include accurate data grounding, constrained generation, and fallback paths for escalation to human agents. Additionally, maintaining transparency, traceability, and alignment with brand standards is crucial. Effective voice AI systems incorporate best-in-class speech-to-text technology, intent detection, and retrieval-augmented generation to enhance accuracy and reliability. Ongoing monitoring and human-in-the-loop feedback are essential for refining these systems and maintaining trust, with safety being a continuous quality assurance focus.

Aircall, a leading AI-powered voice platform, significantly reduced its transcription time by 95% through the integration of Gladia's speech-to-text API, transforming its operations in the contact center industry. By adopting Gladia's solution, Aircall improved the speed and accuracy of its transcriptions, processing over a million calls weekly across multiple languages, which enhanced customer service capabilities such as searchability, AI-generated insights, and agent coaching. The decision to partner with Gladia was based on its high performance, speed, developer-friendly API, and cost-effectiveness. This integration not only improved operational efficiency but also laid the foundation for future advancements in real-time transcription and AI voice agents, underscoring Aircall's commitment to leveraging AI for smarter customer communication solutions.