How Pureverse chairlady Lina Chong is teaching AI to listen in Cantonese

A call of that length is less a record than a reminder of how much some students were carrying into the exams, and of why the agent is designed never to hurry a caller while staying alert to the moment a conversation should pass to a human. Many also came back. Where repeat calls could be counted, around a quarter of callers (25.3 per cent) rang again. For Lina, that suggests the service was meeting more than one-off questions or passing moments of stress: some students were choosing to return of their own accord.

The Caritas project gave Pureverse practical experience of running a Cantonese voice service in a Hong Kong organizational setting. Drawing on its local team and linguistic background, the company focuses on how questions are phrased, when clarification is needed and how a conversation should lead to a useful next step. For enterprise clients, that can include collecting information through dialogue, preparing structured summaries and handing context to the staff responsible for followup.

Pureverse sees the same groundwork applying wherever organisations need to hold large numbers of spoken conversations in Cantonese: employee-wellbeing lines for corporates, after-hours enquiries for insurers and healthcare providers, regular check-in calls for elderly-care services, and front-line hotlines that need to sort and support callers before a human picks up. Organisations that already run their own apps or call centres can build the agent in through an API, rather than send callers to a separate service.

Lina is adamant, however, that further expansion will come only as these first models are fine-tuned to meet defined targets. She is prepared to make only those claims about Pureverse that the data can support.

Two moments told Lina she was on the right track, she says. The first came when an elderly man living alone told the voice agent he was happy because, at last, someone was willing to listen to him. The second was the day her team first heard the Cantonese prototype speak, fluently and naturally, rather than in the robotic voice associated with automated systems.

At Pureverse, Lina says, the test for every product and funding choice is: “Does what we’re doing genuinely serve the person calling?” Her philosophy rests on a Chinese saying: “Roses given to others leave a lingering fragrance on your hands.” For Lina, here as in everything else, helping others is self-fulfilment, not sacrifice.