Assisto Technologies’ iAssist integrates with watsonx.ai to streamline call analysis and boost customer satisfaction
People often blend English and regional languages in conversations, especially in countries such as India, and Assisto Technologies saw an opportunity to better support such natural customer interactions. The company envisioned an AI-driven voice solution that could keep pace with high volumes of multilingual interactions while also analyzing the sentiment and emotion behind every conversation.
To make this vision a reality, however, Assisto needed to move beyond traditional approaches to customer support and adopt a solution capable of transcribing, interpreting and analyzing code-mixed speech at scale. Its goal was to help enterprises access real-time insights, deliver more personalized experiences and support smarter, data-driven engagement across every voice-based touchpoint.
To meet these goals, Assisto and IBM collaborated to integrate IBM® watsonx.ai® and IBM Granite® large language models (LLM), fine-tuned by Assisto, into Assisto’s conversational AI platform, iAssist. Together, Assisto and IBM designed a more human-like voice bot capable of understanding and responding to complex, code-mixed inputs across multiple languages.
Assisto’s in-house automatic speech recognition (ASR) and text-to-speech (TTS) systems accurately transcribe and vocalize multilingual speech. Watsonx.ai enables real-time sentiment and emotion detection, while fine-tuned Granite LLMs enhance response quality in code-mixed interactions. The iAssist platform also autogenerates concise call summaries and supports advanced text analysis for contextual understanding.
To further empower customer support teams, an analytics dashboard built with IBM Cognos® Analytics aggregates insights—such as issue trends, sentiment patterns and agent performance—into a unified, easy-to-use interface.
This comprehensive approach allows Assisto to harness the power of AI to transform unstructured audio data into valuable business intelligence.
The collaboration between Assisto and IBM delivered measurable improvements in customer satisfaction and engagement. The enhanced iAssist platform helped streamline operations and deliver more meaningful customer interactions. Key outcomes include:
- 60%–70% reduction in time spent on manual transcription
- Up to 50% cost savings through intelligent automation
- 20%–30% increase in customer satisfaction scores
- 15%–20% improvement in customer retention and engagement
These gains have empowered businesses across banking, telecom, insurance and retail to respond more effectively to multilingual customer needs while also delivering more personalized, emotionally intelligent customer service. With real-time insights at their fingertips, customer support teams can make faster decisions and enhance every interaction.
The upgraded solution is now available across global markets—including the U.S., Europe and Asia Pacific. Assisto plans to further enhance its AI voice platform by supporting more regional languages and expanding emotional context recognition capabilities, helping businesses deliver more intuitive and inclusive customer experiences.
Assisto Technologies (link resides outside of ibm.com) is a deep-tech AI company specializing in multilingual communication intelligence powering speech, voicebot, OCR, and translation solutions for defense, govt. and enterprise sectors.
Founded in 2021, the company’s flagship product, iAssist, is a unified NLP f stack offering multilingual communication intelligence designed to enhance customer engagement and operational efficiency. With a small but smart, highly innovative team, Assisto delivers cutting-edge AI solutions to Empower businesses with seamless, human-like AI voice & language solutions & help organizations scale and innovate.
Legal
© Copyright IBM Corporation 2025. IBM, the IBM logo, Cognos, Granite, and watsonx.ai are trademarks or registered trademarks of IBM Corp., registered in many jurisdictions worldwide.
Examples presented as illustrative only. Actual results will vary based on client configurations and conditions and, therefore, generally expected results cannot be provided.