Hyperlab’s AI enables better telco customer service
By Anushia Kandasivam November 9, 2017
- Chatbots assist with customer service and sales
- High level of automation still requires human effort
SERVICE design firm Hyperlab has expanded its chat app services to help telecommunications companies leverage on the benefits of artificial intelligence (AI) and natural language processing to improve customer experience.
Launched last year, Hyperlab has seen positive customer gains with big-name brands such as Maxis, U Mobile, Hong Leong Bank and Cerebos coming on board to utilise chatbots that operate in several languages
Hyperlab was also recently awarded an innovation and commercialisation grant from PlaTCOM Ventures to extend its language services across Southeast Asia.
The chat app basically allows users immediate connection and conversation with brands when they are seeking information or help. Users can access the chat app by actively clicking on a chat option button on a website or by clicking on an interactive Internet advertisement.
The multiway and asynchronous nature of chat means that the application has some substantial technology behind it – natural language processing that can be customised for each brand and AI that learns the unique needs of each brand and its customers.
To easily understand what the AI does, says Hyperlab co-founder Vic Sithasanan, one should think of it as a cognitive virtual agent that, like a human sales person, can help customers do things and understand products. The AI is not hard coded but rather conversational – it does not use auto responses but understands what the customer is saying and responds accordingly.
Over the past few months, Hyperlab has focused on telcos to enable them to leverage on this AI to improve customer experience, enabling a few telcos to roll out customised chat apps to help with sales and customer service.
Helping customers make decisions
The latest collaboration is with Tune Talk to recruit new users with an advertising product that programmatically targets users and starts a conversation with them on third-party websites about Tune’s new +Vibe plan.
Basically, Hyperlab has integrated its chat app into Internet banner ads so that customers and potential customers can interact with the ad. “This means the ad display turns from an ad into a retail space where potential customers can ask the AI questions about the product,” explains Hyperlab co-founder Chris Greenough.
Hyperlab is the only service provider in the world to integrate a chatbot in a banner ad in this manner. Tune Talk is its second customer for this service; the first was Astro Go Shop. For the latter, the ads were linked to Astro Go Shop’s product catalogue so that the AI is able to direct customers to specific products they ask for, making the retail experience fully interactive.
Besides this, Hyperlab helped launch U Bot for U Mobile in early September. A chatbot that answers customers’ questions on their instalment plans, U Bot has already exchanged thousands of messages with customers, who engage with the chatbot for an average of two minutes.
According to Lee Ping Han, senior manager of Integrated Digital Platforms under U Mobile’s Marketing Division, the natural language processing and AI capabilities of the chat app influenced U Mobile’s decision to work with Hyperlab, as well as the latter’s local knowledge.
“Because Hyperlab is local, we were able to work closely with them to understand our requirements and business, and our target audience’s mindset and behaviour,” he says.
U Bot is currently in a trial stage and only featured on U Mobile’s main Malaysian website. It is a fairly basic bot that helps customers understand instalment plans and recommends plans to them. Lee says that because U Bot is U Mobile’s first attempt at training a chatbot, the telco is trying it out to see if it works before considering a full-fledged investment.
If U Bot does succeed, U Mobile plans to expand its capabilities and possibly feature it on different channels.
“For it to succeed, it has to be a genuinely helpful platform for our customers and the information it provides must be as accurate as possible. Ultimately, it should provide all the information customers require and take away the need for customers to explore our digital platforms for content and services by themselves,” reveals Lee.
One more telco chatbot, Fuzzy the Star was launched in October for ookyo, Maxis’ new pure-play digital brand. Fuzzy is deeply integrated into ookyo’s service, able to answer thousands of questions in two languages – English and Bahasa Malaysia. Fuzzy is currently available to customers as an app and on the web, and a messenger version will soon be launched to recruit new users.
“What we are trying to solve is the complication that the telco’s potential customers face when it comes to choosing a product,” says Vic. Phone plans, for example, tend to be difficult to compare and most people are not sure what they are getting for their money.
“If you speak to a sales person, it is easy to decide which plan to buy because they will ask you a few key questions and recommend a suitable plan. The whole idea of using the cognitive virtual agent is to take that experience and make it digital,” he says.
As important as the front-facing service is the data gathered at the back end. Hyperlab is able to take all the conversational data – what people are asking and how they are asking it – to the telco’s media agencies to improve their search engine optimisation practices and inform the future marketing of their products.
More human than humans
Though AI can solve myriad problems and streamline work, Vic and Greenough say that one of the big challenges Hyperlab experiences with its customers is the gap in expectations. Interestingly, people expect much more out of the AI than it can do.
“The expectation is that AI can do anything, that it can read their whole website and learn by itself. But it can’t do this; you have to teach it,” says Vic.
Greenough adds that people tend to be harder on bots than they are on human-run chat boxes, expecting the bots to provide a wide variety of responses when human-run chats tend to use copy-and-paste canned responses.
This paradox is illustrated by Lee, who says that U Mobile wants U Bot to have a ‘human’ touch and “not sound like a robot”.
Greenough says that with the tech currently available, Hyperlab does write more colourful responses and puts some ‘personality’ into them. “You can definitely do this [have a variety of responses] better at scale with AI,” he says.
Greenough explains using one of the chatbots it has created, which has thousands of possible responses in its database. Digesting all the data it receives and getting insights from it, growing the AI’s units of knowledge and improving its accuracy require human involvement from human data scientists and linguists.
“At this point, you can’t automate everything. It’s a challenge for enterprises to understand that they can’t just install and disappear. There is a lot more value to get out of the AI if you are willing to commit to working on it,” he says.