Salesforce Launches Native Speech-to-Action AI
Salesforce unveiled 'Speech Invocable Action,' a new native capability that allows users to trigger complex workflows using natural spoken language. The feature translates speech directly into executable actions within Salesforce's Flow engine, enabling tasks like updating compensation records or generating benefits reports by voice. The system is designed with full auditability to address compliance concerns in sensitive HR and finance operations.
- This feature is part of Salesforce's broader "Agentforce" strategy, an AI platform that has seen significant customer adoption, with Salesforce reporting over 18,500 deals by the third quarter of fiscal year 2026. - The "Speech Invocable Action" was developed by the Agentforce Speech Foundations team, which includes engineers Yaheli Salina, Karthik Prabhu, and Omri Alon. The team utilized AI-powered development tools to shorten the architecture exploration and development time by an estimated seven to eight weeks. - A key design principle is keeping all audio data within the Salesforce trust boundary, processing it through native platform services to enhance security and data residency, a direct contrast to solutions requiring third-party speech-to-text services. - While the feature is new, broader Agentforce pilot programs have shown productivity gains of over 20% in sales development workflows, and some e-commerce clients report AI agents resolving 30% of customer order requests without human intervention. - The underlying Agentforce Voice capability uses a multi-model approach, employing speech-to-text models like Deepgram, a reasoning engine based on OpenAI's GPT models under a zero-data retention policy, and text-to-speech models from providers like ElevenLabs. - This native voice function integrates directly with Salesforce Flow, allowing voice commands to trigger, kick off, and modify complex, multi-step automations and workflows that can interact with records across the platform. - Competitors like Zoho CRM and Microsoft Dynamics 365 approach voice differently. Zoho integrates with external assistants like Alexa or triggers calls via webhooks and APIs, while Microsoft uses voice primarily for hands-free navigation within specific apps like Dynamics 365 Guides or connects its Customer Voice survey tool to workflows via Power Automate. - The capability is part of a larger trend in enterprise software where voice becomes a primary input for triggering automated processes; according to a 2024 Salesforce report, 78% of service leaders plan to integrate AI voice agents into their operations within two years.
