Tresic picks Speechmatics after Google and Deepgram evaluation
Tresic has chosen Speechmatics to supply the speech-to-text engine underpinning its Intelligence Cloud platform, having run a competitive evaluation against Google and Deepgram. The Colorado-based vendor builds conversation intelligence on top of voice seats that communications and managed service providers already sell, meaning customers interact with their existing provider's brand throughout and have no direct relationship with either Tresic or Speechmatics.
The selection was driven by two requirements: transcription accuracy on difficult real-world audio, and the ability to run entirely within Tresic's own infrastructure. Because partners white-label the platform, conversation data must remain inside infrastructure that partners can account for to their own customers. Tresic operates two on-premises passes of the Speechmatics engine, one optimised for speed and one for accuracy.
Why accuracy became a compliance question
Tresic frames PCI redaction as the foundational layer of its platform, which elevates transcription from a quality-of-life feature to a compliance control. A card number spoken aloud can only be masked if the underlying transcript is correct; a single mis-heard digit means sensitive data is left exposed. The audio involved makes the problem harder: calls from vans, depots, shop floors, and open-plan offices, spoken in a wide range of accents with no scripted input.
Katy Wigdahl, chief executive of Speechmatics, said that diarisation and accuracy carry more weight in this deployment than they would on cleaner audio, and noted that running inside Tresic's own infrastructure carried no accuracy penalty. "The version that works is the one nobody has to think about," she said.
Kevin Nethercott, co-founder and chief executive of Tresic, said the choice of speech layer was shaped by the company's channel model. "Our partners have spent years earning trust with these customers, and it is their name that goes on this. Speechmatics does one of the hardest parts of the job and takes none of the credit, which is the same arrangement we offer our partners."
The platform is built on vCon, the open IETF-track standard for conversation data, which keeps conversation records portable and readable outside the platform. Conversation data is separated by partner, customer, and location.
Market context
The deal reflects a broader structural trend in enterprise AI through 2026: AI capabilities reaching businesses through incumbent software vendors rather than as standalone deployments. Gartner's positioning of AI in the Trough of Disillusionment supports the view that enterprise adoption is increasingly mediated by familiar brands rather than net-new AI products, a dynamic that benefits embedded-layer vendors like Speechmatics and white-label platform providers like Tresic.
Speechmatics competes in a speech-to-text market that includes hyperscaler offerings from Google, Microsoft, and Amazon alongside specialist providers such as Deepgram, AssemblyAI, and Rev. The on-premises deployment model is a meaningful differentiator for regulated-industry buyers: sectors governed by PCI DSS, FCA conduct rules, or HIPAA face data-residency constraints that cloud-only transcription vendors struggle to satisfy. Speechmatics says it processes more than 500 years of audio globally each month across more than 55 languages, a scale figure that supports its positioning as a foundational infrastructure layer rather than an application-level product.