Bot Support: Press one for personalised chatbots

This article first appeared in Digital Edge, The Edge Malaysia Weekly, on October 24, 2022 - October 30, 2022.

"Personality is a big aspect of bots that are built because we believe that it’s a direct representation of the brand itself.” - Rashid

-A +A

Most of us would have encountered a chatbot at least once, whether it while making a restaurant reservation through social media or calling a bank to make enquiries. Some have a love-hate relationship with bots, especially the tedious process of having to press command numbers on their phones and having to wait for an operator to get on the line.

This annoying communication gap between consumers and businesses resonated with Rashid Khan and his partners, prompting them to set up Yellow.ai seven years ago. The San Mateo-headquartered start-up wanted to provide consumers with personalised experiences when talking to digital enterprises, but it has since changed to become a total experience platform.

“We help enterprises automate their support function, sales function and internal automation for employees, all majorly through chat and voice bots,” says Rashid, who is also the chief product officer at Yellow.ai.

“Every quarter since 2021, Yellow.ai has handled more than four billion conversations on the platform.”

Yellow.ai has two types of chatbots: a text-based one that is compatible with social media messaging platforms and a voice chatbot for phone calls. Most of the time, chatbot systems across platforms are siloed, burdening returning customers with having to start afresh by explaining their enquiries again whenever they contact the same brand through a different platform.

Built as an omnichannel platform, Yellow.ai solves this quandary by merging all customer information from various platforms of a brand into a unified system. The system will then create a unique profile of each customer that contains all information and interactions with the brand.

For example, if a customer previously contacted a brand through WhatsApp and is doing so now through Facebook Messenger, the system is able to display both interactions in front of the agent, granting both the chatbot and the agent behind the screen a consolidated view of all previous interactions to provide better context.

“By [providing unified information of customers], I think we’ve been able to scale to more than 1,000 clients across the globe over the last few years,” asserts Rashid who is also the recipient of the Forbes 30 Under 30 Asia award in the Enterprise (Technology) category.

Rashid shares that his clients have seen their operational excellence increase after implementing Yellow.ai’s technology into their customer experience (CX) strategy, including reduced customer support costs and the cost of handling a ticket.

Most importantly, the synchronous system has seen an increase in customer satisfaction as well. “You essentially are able to provide to your consumers [by] always being available on channels where [customers] like to interact with you,” adds Rashid.

In August, Yellow.ai took a step forward by developing the DynamicNLP platform to skip the taxing natural language processing (NLP) training for chatbots. The platform provides a pre-trained model with unsupervised learning that allows the chatbot to go live in minutes instead of months with 97% accuracy.

Enhancing customer experiences

With 1.4 billion chatbot users worldwide, this technological advancement does not only ease brands in their daily business activities but also has the potential to unlock digital inclusivity. The application of voice chatbots includes search engines such as Google Voice Search, Amazon’s Alexa and Apple’s Siri, to name a few.

According to a study by Ogilvy Malaysia, 34% of Malaysians use voice search and the biggest users were from the bottom 25% income segment. This may be due to the convenience of being able to directly command the search engine or messaging platform instead of scratching their heads trying to figure out what to type.

Another barrier that needs to be overcome when it comes to chatbots is the adaptation to a local language that is more familiar to the target audience of a brand. For some languages, the adaptation has been much faster because chatbots are dependent on the data that they are fed.

“Chatbots that mostly converse in English are able to sound like an Englishman or an American,” says Rashid, adding that Yellow.ai has also been actively working with Microsoft and Google in India to get their chatbots to grasp Hindi.

In order for the artificial intelligence (AI) to learn a new language or accent, the team requires multiple people from various demographics within the community to read a total of ten hours of data and record the way that they speak to generate a seamless speaking experience with the chatbot.

“Even if brands were to launch in a particular state or a particular region where a particular dialect is more prominent, we do it with multiple people to be able to form a more homogenised language output in some sense,” says Rashid.

A step ahead from automated interactive voice response (IVR), which requires consumers to press command buttons on the phone before receiving responses, CX solution providers have introduced conversational IVRs that allow consumers to directly convey their enquiries to the chatbot and it will immediately reply with a favourable response.

“Conversational IVR actually has the ability to understand what the user wants and map it to one of the existing options in a much faster way,” says Rashid.

Considering that each brand has a specific target audience that they reach out to, chatbots can be customised to fit the brand’s personality and interact with consumers better. “Personality is a big aspect of bots that are built because we believe that it’s a direct representation of the brand itself,” says Rashid.

Rashid explains that there are various levels of personalisation with conversational IVR and voice AI in general. The first step will be identifying the tone of approach when framing sentences and responding to questions, whether it is a formal or casual approach. This also includes the usage of slangs that may be relevant to the target audience.

Next, brands may also consider the speed of response depending on the target audience. For example, if a brand is targeting senior citizens, the voice AI’s vocal response has to be slower than usual to give seniors more time to digest the information. Accent also plays a big role in making users comfortable with the ongoing interaction with the chatbot.

Rashid asserts that it is essential for brands to define their personalities through the aforementioned steps. “Brands want to invest in unique sounding voices and interaction — that’s really the goal. The project [to create personalised conversational IVR] doesn’t take a lot of time to complete, but from a strategy perspective, it’s a part of their CX transformation for most organisations.”

“We’ve also started seeing clients creating very persona-specific personalities,” says Rashid. An example is PLUS Malaysia Bhd, which invested in a persona-specific chatbot, PUTRI (PLUS Texting Realtime Interface), the first highway customer chatbot.

Rashid advises brands who want to embark on their CX strategy to understand users’ intention and look at the operational metrics there. Next, brands can start involving a partner who can help navigate some of the CX strategy challenges and bring in platforms such as Yellow.ai to solve a lot of those challenges.

“[After that], move on to the next stage, which is largely measuring how things are happening, how they’re improving and so on,” suggests Rashid.

Combining the best of both worlds

Although we are accustomed to chatbots picking up our calls or answering our enquiries online, we would always prefer to speak to an agent or customer service representative directly. Rashid believes that with the advancement of chatbots, which will most likely be used more to answer user enquiries, chatbots and human customer service representatives can complement each other.

Artificial intelligence is capable of managing first-level questions and understanding the issues that customers face such as rescheduling flights or checking the balance on a membership card, but it will not be able to show empathy like a human.

Some situations, such as at an insurance call centre for medical or personal accident claims, will require a customer service representative to speak to customers directly after traumatic incidents. Meanwhile, the omnichannel platform will allow chatbots to immediately transfer information and context to the customer service representative handling the customer.

“Wherever the AI [chatbot] gets stuck or wherever you need to bring in a customer service representative for more empathy, the platform basically enables a seamless handoff between AI and the representative,” says Rashid.