- Users don't care about network metrics — they want to watch their videos
- Hutchison Telecom Hong Kong is shifting from traditional network metrics to user-centric experience monitoring
- AI-driven tools help resolve problems across OTT services and roaming partners
FUTURENET WORLD, LONDON — At 3 Hong Kong, even when consumer complaints aren't the telco's fault, they are the telco's problem.
As is the case elsewhere, when customers have problems with YouTube videos, music streaming or a roaming partner, they blame the network. The stream might be buffering or lagging because of a problem on the streaming provider's network or server configuration, but the consumer complains to 3 Hong Kong, and the telco gets the blame.
"We realized that the way we look at customer experience needs to change," said Damien Leong, CTO for Hutchison Telecom Hong Kong, which operates the 3 Hong Kong telco service. Leong spoke in a chat on the conference stage with Vijay Sivaraman, co-founder and CEO of Canopus Networks, a provider of real-time network observability and user experience tools.
Leong said, "My team will talk to me about throughput or dropped call rate or the bit error rate, and so forth." But these metrics are from the network perspective, rather than the user perspective.
"The customer doesn't care about network metrics. They care about whether the resolution of their YouTube video is reduced or they experience a buffer," he added.
Customers blame the network
Thus, the operator needed a user experience metric, rather than a network metric, to gauge how well they are serving their customers.
"Unfortunately, we have to adopt responsibility for the whole end-to-end chain. It's not just our network," Leong said.
For example, Leong said his team diagnosed a problem for a senior colleague whose music app was buffering. At first, Leong's team looked at wireless RAN coverage, but that team said coverage was excellent.
After two days of investigation, Leong's team discovered that the over-the-top (OTT) music provider had brought down their content delivery network (CDN) for maintenance and was serving music from Japan.
"I asked my colleague if she was listening to Japanese songs. She said, no, it was a Hong Kong song," Leong said.
In a second example, Leong saw that Google Maps worked well on one phone, but not another. The engineers returned an answer within two hours, discovering three Google servers had an issue and the phone with problems was connecting to one of the problematic servers. "We raised a ticket to Google and Google fixed it," Leong said.
Agentic AI to the rescue
Hutchison Telecom Hong Kong deployed agentic AI from Canopus to go beyond a reactive workflow. Instead of addressing problems only after a customer complains, the telco adopts a proactive workflow, detecting outages and other problems before customers complain and automatically sending a ticket to the OTT provider to resolve the problem, Leong said.
The telco's customer service team is working on developing an outage board for OTT services, so that if a customer calls in about a complaint, the company will already know about the problem. "We can deal with customer complaints better," Leong said.
Similarly, when a Hong Kong consumer has a problem with a roaming provider, Hutchison Telecom Hong Kong can inform the partner more quickly and get the problem resolved.
Looking forward, the operator would like to be able to use agentic AI to offer hyperpersonalized upsell services. For example, if a customer in a stadium at a big sporting event is experiencing network difficulties, the agentic AI might automatically offer to sell the customer a higher tier of service, Leong said.
This shift in responsibility — from infrastructure metrics to perceived user experience — requires more than just new tools. It demands changes in organizational culture, customer service strategy, and external partnerships. Engineers must be retrained to think in terms of service outcomes, not just technical performance. Hutchison Telecom Hong Kong may also need to revisit service-level agreements with OTT and roaming partners to reflect shared accountability for the end-user experience.
Putting AI to work for customers
Like Hutchison Telecom Hong Kong, several communications service providers are adopting strategies using AI and machine learning to enhance the customer experience.
Deutsche Telekom is working with Uneeq, a company based out of Auckland, New Zealand, which has developed a “digital human” — a CGI-based character that customers can talk with to get problems resolved. “Max,” the digital human, has a 10% better success rate at solving problems than a chatbot, handling about 10,000 queries a month and growing.
Vodafone has implemented machine learning techniques to anticipate and rectify network anomalies before they impact customers. Collaborations with platforms like ServiceNow have enabled Vodafone to automate service management, providing faster and more accurate responses to customer inquiries.
And AT&T has developed the ProActive Customer Care (PACE) system, which utilizes machine learning to identify and address individual service issues without waiting for customer complaints. This approach has reduced the volume of support calls.
But using AI to troubleshoot customer experience requires changes to the organization and business process, Leong said.
"In most telcos, you don't have an end-to-end team, or if you have an end-to-end team, it will be a planning function, not an operations or performance function," Leong said.
So, to better manage customer experience, the telco built an end-to-end team. "They don't have an excuse anymore. They can't tell me 'no fault found' anymore. They have to pinpoint whatever the root cause is, so we can manage user experience in a good way," he concluded.