June 26, 2026
Sanas Opens Its Speech AI to Developers and Small Businesses
Sanas has expanded its real-time speech platform and opened it to developers and small businesses for the first time.
The Palo Alto firm built its name on Universal Accent Translation, which transforms a speaker’s accent in real time without altering their tone or meaning. Now Sanas extended that work with wider language support, a new analytics layer it calls Speech Intelligence, and a self-service route designed for teams that have never run an enterprise procurement cycle. An early version of this technology Firstsource deployed to remove accent barriers in its contact centres.
Language Translation Now Covers 30+ Languages
Language Translation covers more than 30 languages, with automatic language detection and low latency on live calls. Sanas builds and runs its own speech models rather than leaning on third-party systems, which it says gives it tighter control over performance and a more consistent result across deployments. Universal Accent Translation continues to work in both directions, adjusting how each side of a conversation sounds to the other while keeping the speaker’s expression intact.
One of Sanas’s largest customers, UnitedHealth Group, has delivered more than $2 million in savings and cut complaints about how well agents could be understood by 70%. “When you’re serving millions of customers, even small communication challenges can have an impact on customer outcomes and operational performance,” said Marius Maree, SVP of Consumer Operations at UnitedHealth Group.
Sanas counts UnitedHealth Group, Comcast, Cigna, Vanguard, American Express, Wells Fargo, Wyndham and Robinhood among its enterprise users, and runs at scale through BPO partners including Teleperformance, Alorica, Concentrix, Foundever, iBex, iQor and TaskUs. Alorica’s relationship with the company runs deep enough that it joined Sanas’s $65 million Series B funding round as a strategic investor.
Real-Time Conversation Analysis
The second part of the announcement is Speech Intelligence, an analytics layer that reads conversations while they are still happening. It handles real-time sentiment detection, fraud and abuse signals, compliance monitoring and PII redaction, and runs on-device so teams can act on sensitive calls without sending data to outside systems. Sanas argues this sets live analysis apart from the after-the-fact review most contact centres rely on. “If you are only analyzing conversations after they end, you are missing the moment where outcomes can change,” said Shawn Zhang, CTO and Co-founder of Sanas.
The third part takes Sanas out of the large-enterprise lane entirely. A new self-service experience lets individuals and small businesses sign up directly, test the technology on live calls and judge it for themselves before committing. Alongside it, a developer platform and SDK let teams embed Sanas’ speech tools into their own products, collaboration apps and AI applications without rebuilding their workflows or retraining staff.
Cresta CEO and Co-founder Sharath Keshava Narayana said the company’s focus has stayed narrow. “We built Sanas to fix the part of communication that actually breaks,” he said, arguing that a cleaner speech signal improves every downstream system that depends on it, from transcription accuracy to automated routing.
