IBM has moved to strengthen the voice capabilities of watsonx through a new partnership with speech-AI company Deepgram, announced in late February 2026. The agreement brings Deepgram’s speech technology to watsonx Orchestrate, IBM’s platform for building and managing AI agents within enterprise workflows, and marks IBM’s first collaboration with a dedicated voice technology provider.
What the partnership delivers
Under the deal, Deepgram’s speech-to-text and text-to-speech technologies will be embedded directly into the watsonx Orchestrate Agent Builder. The companies said the integration will provide enterprise-grade transcription, real-time captioning, and natural-sounding voice interactions for digital agents, allowing IBM clients to build voice agents and voice-enabled workflows on top of a real-time speech foundation that Deepgram has refined over more than a decade.
The move comes as organizations increasingly look to add voice control to AI tools, with advances in large language models making conversational, voice-driven agents more practical. Deepgram’s technology is designed to handle everyday audio challenges such as background noise and varied speech patterns, and supports a range of languages, including several Arabic and Indian variants.
Why voice, and why now
“Voice is rapidly becoming the default interface between humans and technology, and enterprise deployments require a real-time platform that is accurate, low latency, and reliable at scale,” said Scott Stephenson, co-founder and chief executive of Deepgram, in a statement. He added that embedding Deepgram inside the watsonx Orchestrate Agent Builder lets IBM customers build voice agents on a real-time foundation built and refined over more than a decade.
For IBM, the integration adds a specialized voice layer to watsonx at a time when conversational AI interfaces are becoming a standard expectation in enterprise software. “This collaboration aims to help enterprise organizations accelerate their AI initiatives and strengthen IBM’s open ecosystem, bringing choice and cutting-edge voice technology to partners and customers,” said Nick Holda, IBM’s vice president of AI technology partnerships. For Deepgram, the deal expands access to IBM’s enterprise client base.
Initial use cases
The companies said the integration will initially target customer support, call analytics, and voice-driven data entry, with healthcare and finance cited as priority sectors. These are areas where accurate transcription and low-latency responses matter most, and where the volume of voice interactions is high enough to justify automation.
Limitations and what to watch
As a vendor announcement, the partnership’s stated benefits reflect the companies’ positioning rather than independent evaluation, and real-world accuracy and latency will depend on deployment conditions, audio quality, accents, and domain-specific vocabulary. Voice agents in regulated sectors such as healthcare and finance also raise additional requirements around privacy, consent, recording, and compliance that the underlying speech technology does not resolve on its own. As with any agentic system, organizations should pilot carefully, measure transcription accuracy on their own data, and keep appropriate human oversight in high-stakes workflows. Related coverage on this site examines how enterprise agent frameworks are built and governed.
Further detail is available from the IBM newsroom.
