SMU-21-0004A__Bus_Uncle_A_.pdf

SMU-21-0004

This case was written by Wong Yuet Nan, Associate Professor Hoe Siu Loon and Dr. Jovina Ang at Singapore Management University. The case was prepared solely to provide material for class discussion. The authors do not intend to illustrate either effective or ineffective handling of a managerial situation. The authors may have disguised certain names and other identifying information to protect confidentiality. Copyright © 2021, Singapore Management University Version: 2021-05-21

BUS UNCLE CHATBOT – CREATING A SUCCESSFUL DIGITAL BUSINESS (A)

Is the bus here yet? Oh, I wish I could just ask someone so that I know how much longer I

have to wait!

September 2016 – The idea to create Bus Uncle sparked while Abhilash Murthy was waiting for

Bus 65 along Singapore’s Orchard Road. After putting in a long day at work, Abhilash was tired

and hungry. He simply wanted to go home. But his bus was nowhere in sight despite him waiting

for 25 minutes.

“When is the bus coming?” was the simple problem that he wanted to solve. Despite the

availability of several bus apps, he did not want to use any of them, as “there was too much

information”. Neither did he want to walk to a bus stop where there was an electronic bus board

that displayed the bus arrival times. Like the bus apps, the electronic bus boards also displayed

too much information. He simply wanted to ask someone when the bus was coming.

Since Facebook Messenger had just launched chatbots on its platform, Abhilash decided to jump

on the bandwagon and create a chatbot to solve the problem. Trained as a software engineer, he

took just one weekend to hack out a chatbot prototype that he could ping to ask when the bus was

coming. The chatbot prototype was later renamed Bus Uncle and given a persona that typified a

Singaporean bus captain – grumpy, snarky but helpful.

Abhilash launched Bus Uncle to the public on October 23, 2016 on his Facebook page. Despite

little interest in the first four days, by the fifth day, the Singapore public took Bus Uncle by storm

when Mothership, a popular local media, featured the chatbot on its platform.1 In just one day,

Bus Uncle went from having 100 users to more than 8,000 users.

What were the reasons why Bus Uncle was so popular? What could Abhilash do to enhance the

chatbot?

Abhilash’s Background

Originally from India, Abhilash moved with his family to Jakarta when he was three years old,

living there for 15 years before moving to Singapore in 2010.

Abhilash developed an interest in coding from a very young age, and was often seen tinkering

with software. By the time he was 12 years old, he could write programmes in Java2. He added,

1 Jonathan Lim, “'Bus Uncle' is the Most Singaporean Bus Arrival Timing App Ever”, Mothership, October 27, 2016,

https://mothership.sg/2016/10/bus-uncle-is-the-most-singaporean-bus-arrival-timing-app-ever/, accessed March 2021. 2 Java was one of the top three programming languages in the world. It was an open source programming language that was

platform independent and could run on both mobile and desktop devices.

SMU-21-0004 Bus Uncle Chatbot – Creating a Successful Digital Business (A)

2/10

We had a very simple computer [at home] and it was too slow for the games that I was

playing. I would try to find scripts online and edit the scripts so that I could manipulate the

computer through the scripts.

Because he loved programming, Abhilash knew from an early age that he wanted to be a software

engineer. When the opportunity arose for him to pursue his studies in Information Systems at

Singapore Management University (SMU), without any hesitation, he decided to move to

Singapore alone.

Abhilash excelled in his studies at SMU. He even pursued a second major in Advanced Business

Technology to broaden his knowledge of new technology. When Facebook launched its inaugural

Hackathon in Singapore, he and three other friends at SMU signed up to compete and secured a

place in the finals at Facebook Headquarters.

Starting Out at TradeGecko

Fresh out of university, Abhilash started as a Software Engineer Graduate Trainee at UBS, an

investment bank. Only after working at the bank for a year did he decide to move to a start-up,

and joined TradeGecko3, a two-year old start-up that focused on providing a cloud-based

Amazon-like supply chain order and inventory management system for small and medium

enterprises. Given that the team was lean, he was immediately roped in to develop software for

the growing start-up.

TradeGecko was founded by brothers Cameron and Bradley Priest, and Carl Thompson.4 The

three New Zealanders ran the company like a tight ship as they had just secured US$650,000 seed

funding from WaveMaker Labs and Golden Gate Ventures.5 Daily standups were a part of the

software development process. During the standups, team members were expected to give a 20-

second update on their work, what they would work on next, and whether they were encountering

any blockers. Sprints were held every six weeks during which new tasks would be assigned to the

software engineers and designers.

Even though Abhilash worked at TradeGecko for only two years, he thoroughly enjoyed his time

there.

I had so much fun working at TradeGecko. Work certainly didn’t feel like work at all! And I

got to learn so much.

The founders created a culture that was centred on experimentation and they encouraged all

members of the small team to be independent and take charge of their own career paths. Everyone

who worked there was encouraged to have a side project. And they specifically told Abhilash,

“The best path you could take is to leave us and build your own start-up.”

He was exposed to the methods of Design Thinking, Lean Start-up and Agile for developing

software, even though in his words, “the methods employed at TradeGecko did not follow the

textbooks to a t”. Despite having responsibility for developing front-end user interfaces and

3 Cameron Priest, “The Story Behind TradeGecko’s USD 10M Series B Announcement”, Intuit Quickbooks, July 23, 2018,

https://www.tradegecko.com/blog/tradegecko/the-story-behind-tradegeckos-usd-10m-series-b, accessed March 2021. 4 Intuit Quickbooks, “Our Founding Story”, https://www.tradegecko.com/about-us, accessed March 2021. 5 Digital News Asia, “Investors Get Sticky with TradeGecko”, December 18, 2012, https://www.digitalnewsasia.com/sizzle-

fizzle-slow-burn/investors-get-sticky-with-tradegecko, accessed April 2021.

SMU-21-0004 Bus Uncle Chatbot – Creating a Successful Digital Business (A)

3/10

interactions, he had to learn how the whole technological stack was tied together so that his work

output could be seamlessly integrated to the back-end. He also gained an appreciation of the

importance of design, and elaborated,

At TradeGecko, we put a lot of thought into how everything was designed because we wanted

to create an addictive user experience. We even incorporated design into the colour and size

of the buttons!

The Idea to Create Bus Uncle

2016 was the year when chatbots were the rage in the technological industry. All major

technological companies were focusing on chatbots. In April 2016, Facebook released chatbots

on its Messenger platform.6 Google followed suit a month later with the announcement of a new

chat service called Allo.7 And Satya Nadella, Microsoft CEO, said that chatbots would

“fundamentally revolutionise how computing is experienced by everybody.”8

The idea to create a chatbot was sparked on a late evening in September 2016 when Abhilash was

waiting for a bus to arrive. Given that he wanted to ask someone when the bus was coming, the

idea of a chatbot fitted the bill. Talking to himself, he thought,

Wouldn’t it be cool if I had a chatbot that I could ask when my bus is coming?

On the bus ride home, he visualised what this chatbot would look like. Getting excited, as soon

as he reached home, he started working on the chatbot prototype on Facebook Messenger.

Abhilash knew what type of chatbot he wanted to create. He did not want one that gave

uninspiring and robotic answers, or one that responded based on a structured prompt tree. Instead,

he wanted to create a human-like chatbot – a chatbot that he could converse with, as well as one

that had a bit of humour so that users would have a delightful experience interacting with it. He

explained,

Chatbots don't have any user interface. There’s no design element, no buttons and no menu.

You send it a message and it returns one. It’s easy to create a functional and utilitarian

chatbot. Given that chatbots were still very new then, what I’d notice was that when people

speak to chatbots, they generally feel defensive and uncomfortable. I wanted to create

something that makes people feel good and familiar while interacting with it.

Despite having very little sleep that weekend, by Sunday evening, Abhilash had accomplished

his goal and had a chatbot prototype to show his friends and colleagues. The first thing he did

when he arrived at work the next morning was to get his friends and colleagues to test the chatbot

out. While he received great feedback from the people he reached out to, it was his close friend

Chan Yin Yin, who suggested that the chatbot should speak Singlish9 and be renamed Bus Uncle

so that it could personify a typical Singaporean bus captain: grumpy, snarky but helpful.

6 Josh Constine, “Facebook Launches Messenger Platform with Chatbots”, Techcrunch, April 13, 2016,

https://techcrunch.com/2016/04/12/agents-on-messenger/, accessed April 2021. 7 Jennifer Booton, “Google Follows Facebook into Chatbots”, MarketWatch, May 19, 2016,

https://www.marketwatch.com/story/google-follows-facebook-into-chatbots-2016-05-18, accessed April 2021. 8 Khari Johnson, “Microsoft CEO: Chatbots will ‘Fundamentally Revolutionise’ Computing”, Venture Beat, July 11, 2016,

https://venturebeat.com/2016/07/11/microsoft-ceo-chatbots-will-fundamentally-revolutionize-computing/, accessed April 2021. 9 Singlish was a variety of English spoken in Singapore – it included words from the many languages and dialects spoken by the

people who lived in Singapore including Malay, Hokkien, Teochew, Cantonese and Tamil.

SMU-21-0004 Bus Uncle Chatbot – Creating a Successful Digital Business (A)

4/10

Putting themselves in the shoes of Bus Uncle, the two friends imagined Bus Uncle’s likes and

dislikes, his relationships and interactions with commuters, as well as what he would do in his

free time.

We have seen how grumpy bus uncles can become. At times when the buses are crowded, we

have seen them shouting to the commuters to move to the back of the bus. They might also

get grumpy and start yelling at people when commuters take their own sweet time to get on

the bus. At the same time, we have noticed that the typical bus uncle likes to sit down at the

kopitiam (coffee shop in Hokkien) to drink coffee during their breaks or have a Tiger Beer

after work.

As a Singaporean, Yin Yin was well versed in Singlish. She helped Abhilash create a few logical

conversational flows in Singlish while he focused on leveraging artificial intelligence (AI),

specifically Wit.ai10, to create the rules using natural language processing (NLP) to make the

chatbot talk and sound like a human being (see Exhibit 1).

After doing some tweaking and enhancements to the conversational flows, Bus Uncle was ready

to be launched. On 23 October 2016, Abhilash launched Bus Uncle to the public on his Facebook

page. Despite acquiring only 20-30 users per day in the first few days, Bus Uncle became quite

popular when it was featured on Mothership, a local online media platform that was frequently

read by Singaporeans. Within a day, Bus Uncle went from having about 100 users to more than

8,000 users.11

This huge surge of users nearly crashed the web server that hosted Bus Uncle. Abhilash started

getting all sorts of error messages from the server. He knew he had to do something quickly. He

immediately sought help from his colleagues to learn about web server architecting for handling

large amounts of users and data. Within a month, he was able to bolster the chatbot and fix the

technological issues by re-architecting the setup (see Exhibit 2).

Testing and Learning

While he was addressing the technological issues, Abhilash was proactively managing user

relationships and user expectations. Working together with Yin Yin, the two friends shared the

wide-ranging and funny conversations that users were having with Bus Uncle on the Facebook

page to build relationships with them. The conversations with Bus Uncle could be as short as 5

minutes to as long as 30 minutes.

For example, one person told Bus Uncle that his landlord had kicked him out of the house.

“Uncle, my landlord just kicked me out of my house”

Bus Uncle responded: “Where are you now?” (Guffaws)

“I’m so angry and upset.”

Bus Uncle replied: “There, there. Everything will be okay.”

10 Wit.ai was a natural language interface for applications, capable of turning sentences into structured data. It was popular with

chatbot developers as it was easy to make the chatbot interact with human beings on messaging platforms. Wit.ai was acquired by Facebook in 2015.

11 Jonathan Lim, “'Bus Uncle' is the Most Singaporean Bus Arrival Timing App Ever”, Mothership, October 27, 2016,

https://mothership.sg/2016/10/bus-uncle-is-the-most-singaporean-bus-arrival-timing-app-ever/, accessed March 2021.

SMU-21-0004 Bus Uncle Chatbot – Creating a Successful Digital Business (A)

5/10

Abhilash relied heavily on data to make his decisions. He constantly looked for new ways to keep

the users engaged and took every opportunity to communicate with them. He tracked all the

messages from the users to understand the patterns of conversations, so that he could continue to

inject the human element in Bus Uncle’s responses. The Monday joke subscription was a weekly

subscription feature that he introduced. When he decided to get Bus Uncle to tell jokes on a daily

basis, the users told Bus Uncle to stop as they felt that this had become like spam.

The users told him that they wanted Bus Uncle to reply when they said thank you. Thus, Abhilash

programmed the chatbot to say things like (see Exhibit 3 for the definitions of common Singlish

words):

I shy leh.

You happy can already lah.

Okay I love you too.

He also sought feedback about the new features that he was planning to introduce. One important

feature that the user requested was the ability for Bus Uncle to capture the bus stop number. This

led Abhilash to add the image recognition AI feature on the chatbot.

What’s Next for Bus Uncle?

Time had passed quickly. Before Abhilash knew it, December had arrived and Bus Uncle had

been operational for three months. In this time, Bus Uncle had become Singapore’s most popular

chatbot with 20,000 users12 who pinged him regularly to ask for bus arrival times or simply to

chat.

Abhilash certainly had learned a lot in this time, especially how people talked to chatbots and

what they felt when they did so. It was clear that the human element was what attracted users to

engage with Bus Uncle.

He had also received lots of feedback about why users loved Bus Uncle (see Exhibit 4). He

narrated,

People love Bus Uncle because he is witty, sarcastic, funny, informative, and most of all,

local and relatable. When you ask him when the bus is coming, his responses are exactly

what you would expect from a Singaporean bus captain. He responds with lah’s, aiyo’s,

jialat’s, walao’s. He also has a sense of humour. If your wait is longer than 15 minutes, he

might even ask you to lim kopi (drink coffee in Hokkien).13

Bus Uncle had remained a fun side project for Abhilash. It was not generating any revenue for

him even though all of his free time was dedicated to it. He was still operating alone, with

occasional help from Yin Yin, while holding his full-time job at TradeGecko.

There was still a lot more that Abhilash believed he could do to improve Bus Uncle. The chatbot

could not yet understand human sentiments such as frustration, anger, urgency or kindness. What

12 Cherise Fong, “Top 10 Chatbots of Our Times”, Makery – Media for Labs, December 5, 2017,

https://www.makery.info/en/2017/12/05/top-of-the-bots/, accessed April 2021. 13 Interview with Abhilash Murthy on March 2, 2021.

SMU-21-0004 Bus Uncle Chatbot – Creating a Successful Digital Business (A)

6/10

else could he do especially when his dream was to maintain Bus Uncle’s position as Singapore’s

most popular chatbot?

SMU-21-0004 Bus Uncle Chatbot – Creating a Successful Digital Business (A)

7/10

EXHIBIT 1: SCREENSHOTS OF BUS UNCLE’S RESPONSES

Source: Abhilash Murthy

SMU-21-0004 Bus Uncle Chatbot – Creating a Successful Digital Business (A)

8/10

EXHIBIT 2: BUS UNCLE TECHNOLOGICAL ARCHITECTURE – DAY 1 VS. DAY 30

Source: Abhilash Murthy, “Presentation to SMU Students”, February 22, 2021.

SMU-21-0004 Bus Uncle Chatbot – Creating a Successful Digital Business (A)

9/10

EXHIBIT 3: COMMON SINGLISH WORDS AND THEIR MEANING

Singlish Meaning

aiyo An expression of surprise or it could apply to annoyance

as well.

boleh It means can in Malay

chope To reserve or hold something for someone.

lah A suffix used to emphasise the sentence or a prior word.

leh An expression when one is not sure about something or

when one is modest.

jialat A word that is used when nothing is going in one’s favour

or way.

sabo A word that is used to play a practical joke or cause harm.

sian It means tired in Hokkien. It can also mean boredom or a

lack of enthusiasm.

siao It means crazy in Hokkien.

walao An exclamation that means annoyance, disbelief,

frustration or surprise.

Source: Honeycombers, “The Top Singlish Phrases You Must Know to Chat Like a Local”, Honeycombers,

November 6, 2019, https://thehoneycombers.com/singapore/singlish-101/, accessed April 2021.

SMU-21-0004 Bus Uncle Chatbot – Creating a Successful Digital Business (A)

10/10

EXHIBIT 4: BUS UNCLE SUCCESS FACTORS

Source: Abhilash Murthy, “Presentation to SMU Students”, February 22, 2021.