AI-generated text and voice have significant applications in CX services, with multilingual support now offering the most transformative use cases. Transcom has been actively building its capabilities in AI voice with its T:Translate platform and already has live implementations.

Integrating AI translation into the existing CX ecosystem

T:Translate is an automated real-time translation application using AI to answer customer queries over text channels: emails, tickets, live chat, and messaging. And now in 2023, Transcom has launched AI voice translation.

T:Translate combines in-house developments with GenAI third-party speech-to-text software. The company assessed new niche GenAI, LLM, and specialized voice AI vendors to derive specific UX and feature sets for building a real-time translation engine for voice and text. The goal is to achieve higher UX and quality than standalone translation engines. For example, Transcom is working with a Berlin developer, Parloa, for the speech-to-text and text-to-speech technology. For text specifically, T:Translate is customizable and embedded in existing contact center systems such as Salesforce, Genesys or Zendesk and other conversational AI solutions. For voice, it does not require direct integrations, making voice translation faster and more scalable to implement than text.

T:Translate also has a tailored translation glossary and features for contextual, industry, and brand-specific vocabulary and can also change the tonality, rate, pronunciation, and speaking style. Transcom provides the LLM training upfront using existing knowledge bases, interaction recordings, and sentiment analysis. It also builds the glossary using other resources, such as the brand website and training materials.

The tool allows the agent to handle multiple languages with an average translation accuracy equivalent to a B2 or C1 language level agent or +90%. T:Translate supports ~100 languages and gives a choice of ~400 voices across genders and ages. The company can typically deploy it in six to seven weeks.

Augmented agent

Transcom sees AI translation as one of the transformation tools with the greatest impact, driving improvement in the customer experience, lowering costs while maintaining quality for clients, and increasing the agent experience. T:Translate is part of the overall company approach in offering advisory, technology, and transformation of the client setup through its T:Labs innovation unit. T:Labs is responsible for T:Translate implementation and subsequent training and maintenance of the LLM, focusing strongly on the algorithm’s self-learning from the agents’ selection.

The AI can be trained in different languages, but usually, Transcom trains it in English as the middle language and monitors for translation quality post-implementation. Another service monitors the quality of the AI outputs and optimizes the translations.

T:Translate addresses increasing market expectation to improve operational performance through automation and co-pilots on different process parts, such as call wrap-ups in languages the agent is not fluent in. Transcom positions AI voice as the next-generation solution for these more complex transactions that cannot be redirected to text effectively. Transcom’s experience using T:Translate in combination with these RPA/RDA, intelligent automation, and agent augmentations such as next-best-action achieves TCO reduction, faster issue resolution, and increase in FCR and NPS. In TCO, the company expects to deliver between 25% and 65% reduction when deployed on voice automation to augment live agents. Another benefit is accuracy and compliance adherence; for example, the tool always uses the right wording per the processes. T:Translate is GDPR compliant and does not demand storage of the transcriptions.

AI alternative to traditional multilingual support

Traditional multilingual delivery faces challenges in sourcing fluent language talent that requires higher compensation. When this talent is available, operations have difficulty maintaining effective utilization, with uneven volume patterns between language queues. With T:Translate, one agent can handle multiple languages without any language skills, achieving productivity of ~80-85%, allowing Transcom to run lower-cost locations with a lower headcount. An average hourly saving for a Nordic support role is ~50%, taking the rate to under €18/h.

In 2023, Transcom established an AI translation hub in Cairo, where it has a multilingual capability, including for Nordic languages. It is gradually routing more of its multilingual volumes there.

An example implementation is for a U.S. CPG company that wanted to enter European markets covering multiple languages with small volumes by using automated translation. In Q1 2023, Transcom deployed real-time voice translation for European languages in English and Swedish with fluent speakers from Malaga, Spain, and 100% automated translation in Polish, German, Italian, and Czech within one month and without system integration. The program has a 0% CSAT deviation.

Expansion of T:Translate to the client base

On the development roadmap, Transcom is adding functionality for agents to flag translations for checking and retraining of the AI. It also develops language skills and selects agents with language aptitude in Cairo to become translation experts and AI trainers. It also plans to recruit a linguist to develop the glossaries and technology further. Another area is testing to identify the right voice characteristics for the different support scenarios. With the rapid evolution of the quality of synthetic voice, Transcom looks at significant opportunities for creating custom brand voices and new brand identities. At the same time, certain brands want more robotic-sounding voices to distinguish them clearly from human voices.

The company is now proactively offering the full hybrid model (with bots, AI translation, and interaction shift to text) to its existing clients. It has three projects in development with go-live dates in Q3 2023, and for the U.S. consumer brand, it is adding six more European languages. Target processes are technical support, urgent queries, and handling peak volumes.

Regardless of the technical leap forward, the AI translation will not completely eliminate the need for onshore and nearshore CX support. Regulated processes, higher value interactions, and highly unstructured conversations require cultural and language proximity. An example of this market direction is Transcom’s investment in July 2023 with the acquisition of a German CX services provider, timeframe, adding employees in Germany, Portugal, and Greece.