Chatbots have been gaining popularity over the recent years, with many brands jumping on the hype and launching their own as an additional avenue to engage with consumers. A study by Forrester titled “Chatbots are transforming marketing” also found 57% of companies globally already using chatbots or plan to do so.
While chatbots are all the rage right now, brands should still ensure there is proper planning rather than launching a chatbot for the sake of it. It is also important for companies to not oversell the return on investment (ROI) of their chatbot project, Kanags Surendran (pictured right), managing director and head, digital banking, CIMB Bank, said during the panel at A+M’s Digital Marketing Asia 2018 held in Kuala Lumpur.
It takes a period of time before a new technology becomes mainstream, and if brands do have the bandwidth to test it out, they should proceed, Surendran said. “But don’t start the journey expecting to reduce your call centre volume by 50%, or expect to fully direct consumers away from your website within the next two years, for example. It’s not going to happen,” he added.
According to him, digital accounts for 25% of the bank’s marketing spend. CIMB’s enhanced virtual assistant (EVA) was launched last year to help consumers with basic activities such as funds transfer, bill payments, credit top-up and analysis of expenditure.
The bank chose to take on a transactional route with EVA because the requirements were more straightforward, making the chatbot “easier to train”. However, Surendran said that CIMB is taking steps to personalise EVA through natural language processing (NLP), so customers will not lose interest or trust in a chatbot. He added:
I think the brand recognition and marketing PR value that we generated from the launch of EVA was more than sufficient to bear the cost that went into creating it.
Agreeing with him is Peter Pohlschmidt (pictured centre), head of digital, Malaysia Airlines (MAB), who said that companies should be “mindful” of how they position the chatbot’s ROI when creating one, as well as how they justify its value.
The airline views its chatbot, MHchat, as a transactional bot which allows consumers to make bookings online and check their flight itinerary. In addition to the monetary ROI set for MHchat, MAB also uses it as another service channel for consumers.
“MHchat is primarily still about self-servicing and empowering customers, rather than escalating the transactions to call centres,” Pohlschmidt said. Consumers frequently ask standard questions such as “What’s my baggage allowance?”, and such interactions can be shifted from the phone and Twitter to MHchat, allowing for customer self-service.
Why do chatbots fail?
Despite more than half of companies worldwide currently using or planning on using them, Forrester’s “Chatbots are transforming marketing” study said that most chatbots still fail. According to Pohlschmidt, this is because companies do not set proper internal expectations of what it is trying to achieve with the chatbot. He said:
One thing you need to realise is you’re launching an imperfect piece of technology because it is evolving.
“The bot framework is essentially a cognitive service which is still learning and evolving over time. As such, you need to manage the expectations of internal stakeholders because a chatbot can produce incoherent conversations,” Pohlschmidt said.
Also weighing in on the conversation is Liew Ooi Hann (pictured left), CEO of financial comparison website RinggitPlus.com, who agreed that it is important to keep the context of the chatbot “narrow” to minimise the chances of failure when launching a chatbot.
Despite having limited capabilities, consumers will still know what to expect from chatbots. This is more useful as compared to saying “I have a chatbot to serve customers”, Liew said, which might set a wide perimeter and have consumers expect the chatbot to be able to answer all types of queries.
The primary objective of RinggitPlus.com’s chatbot, which was done in collaboration with RHB Bank, was to improve the onboarding experience for consumers who were applying for financial products on RinggitPlus.com. According to Liew, the bot saw a conversion rate of 60% during the past six months, up from 20%.
The challenges of understanding cultural nuances
Unlike humans, chatbots do not understand colloquial languages such as Manglish (Malaysian english) or cultural nuances. It is important for chatbots to understand Manglish because consumers might tend to use colloquial language when messaging, Liew said. “We need to be able to handle the ‘lahs’ and ‘kthxbye’. If you don’t, you will not be able to capture consumers’ attention and your bot might fail,” he added.
Despite seeing approximately 100,000 conversations a month, Liew said that the volume is “not big enough” for NLP algorithms to work well. While he believes it is important for chatbots to understand Manglish or cultural nuances, Liew said that there is still a need to clearly define to what extent should NLP algorithms be expected to understand those aspects before having a live agent take over.
Meanwhile, CIMB’s Surendran said it would be a challenge to have chatbots learn cultural nuances.
“It is not easy for a chatbot to differentiate a phrase which is typed in Latin characters to realise that it is not English but Bahasa Melayu or even Manglish,” Surendran said. As such, the bank is currently focusing on English to ensure the EVA is working smoothly according to the decision trees it has planned out, before focusing efforts on different languages.
However, Surendran said the bank did take steps to train EVA to understand approximately 2,000 words in English and Bahasa Melayu to ensure it responds appropriately. The bank currently has approximately 200,000 interactions on the chatbot but even with this volume, Surendran said it might take “awhile” for consumers to have real, personalised conversations with chatbots.
“Also, because our chatbot is a separate mobile app, getting them to download the app is also another challenge. We plan to integrate EVA into our traditional platforms by the end of this year, and expect the volume of conversations to increase.
Can companies do without a chatbot?
A chatbot can help reduce the number of transactional activities that employees or call centres have to deal with. According to Surendran, questions such as “How much balance do I have?” or “I forgot my password” do not require assistance from call centres and often end up creating large volumes of calls. If consumers lose their card, for example, they will either want to cancel it or be reissued a new one. And simple conversations such as this, Surendran said, can be resolved by the chatbot.
“These are the things you need to look at and see how we can move consumers into a system that is self service. It will help improve customer experience and the efficiency of the organisation,” he said. Surendran added that having a chatbot is equivalent to creating a new channel, and not being present would be a detriment for the business.
Agreeing with the concept of self-service is MAB’s Pohlschmidt, who said that B2C companies, especially those in the airline industry, need to have self-service automation that empowers consumers because contact centres might not have the resources to manage all enquiries. Transactions on a chatbot is one application of self-service automation, he said.
While the necessity for a chatbot depends on the needs of a company’s consumers, Pohlschmidt said it is still important for companies to at least have the self-servicing framework to more efficiently manage the scale of its customer service operations. Nonetheless, he said that it is still important to manage consumer expectations on what a chatbot can or cannot do, and that MAB would not run anything “mission critical” on a chatbot.