Calling the first-generation iPhone just another phone with a touchscreen does not do it justice. Over the past decade, the device has outgrown its capabilities and expectations, revolutionising the way we communicate with each other and giving birth to a whole new industry.
Chuang Shin Wee, co-founder and CEO of Pand.ai Pte Ltd, likens this to referring to artificial intelligence (AI) chatbots as a simple business tool for customer support. “When you contact a bank’s call centre, you are put through an Interactive Voice Response system, where you have to wait and press a lot of buttons before you can have your problem solved. Chatbots are different because they are able to provide immediate support to customers,” he points out.
“In a way, AI chatbots are software-based, so they are extremely scalable. Even if you have 1,000 people calling in at one go, the chatbot can handle these volumes. Right now, it is unlikely for an organisation to have a call centre that is well-staffed enough to handle such volumes.”
Pand.ai is a Singapore-based AI financial technology (fintech) company that develops and supports AI chatbots specifically for financial institutions, which includes banks, insurance providers and asset management companies. Its name is a play on the Malay word for clever — pandai — and AI.
Pand.ai’s chatbots have been deployed for many other use cases outside of customer support, including recruitment, marketing, human resource (HR) management and staff training. “Clients can deliver a two-minute piece of learning material directly to the sales team’s phones for them to learn certain things [related to financial products] and test their understanding of the materials, and for the sales team to ask questions, which turns the chatbot into a virtual professor. The objective is to have sales professionals become more educated so they can serve customers better and provide higher quality financial services,” says Chuang.
Pand.ai even uses chatbots internally for HR-related matters. Chuang cites the example of claiming one’s travel expenses by answering several questions posed by the chatbot and then uploading screenshots of the receipts as proof of payment. The entire process takes only a few minutes.
From his experience, Chuang notes that many financial institutions are keen on using AI chatbots to generate leads. This is partly due to how the financial sector operates. “Financial products, for the most part, are sold and not bought. Most of the time, when people buy their first life insurance, it is not because they were browsing online and added insurance into their shopping cart. It is because somebody sold it to them,” he explains.
“That is because financial services and products are both very complicated and personalised. There is no best product for everyone. There is only the right product for the right customer. So, there is a fair amount of customisation and understanding that needs to be delved into during the buying process.”
This is why relationship managers and financial advisers play a vital role in the financial ecosystem. There is some degree of interaction that needs to take place before a customer is ready to make a buying decision.
However, most financial institutions hit a roadblock when it comes to their official websites. Chuang points out that most websites serve as a static catalogue of information and navigating through the information lies in the hands of customers, not the financial institution. Customers are left to their own devices trying to figure out which product is right for them.
“We do not see websites as effective tools for converting online visitors into leads. But chatbots allow for more meaningful interactions and
conversations to take place, thereby helping customers pick the right product or, at least, have the right conversations. As a result, when we help clients develop AI chatbots for lead generation, we often find that chatbots are able to outperform traditional websites by up to 10 times,” he says.
Chuang derived his findings from conducting A/B testing (a user experience research methodology) on the client’s website, with and without the chatbot. However, he hastens to add that merely incorporating AI chatbots does not automatically help businesses to generate 10 times more leads.
“That is because if you rely on organic traffic, chances are, the people who land on your website already have a specific purpose for visiting. A chatbot can help, but not by a tenfold magnitude. Where we see the most impact is when the brand runs an advertising campaign, which draws in new customers,” he says.
However, different clients have different needs, warranting different performance indicators, says Chuang. Many factors are taken into consideration when customising and deploying AI chatbots such as the client’s circumstances, types of products and business goals.
“I think it is an oversimplification to say that chatbot AIs are primarily lead generation or recruitment tools. At its core, chatbots are communication tools for companies, externally and internally. Depending on what the client is trying to communicate, the chatbot is built to fulfil that business goal,” he says.
While the prospects are bright for AI chatbots to become the Swiss Army knives of businesses, the technology is still not mature enough to solve all business problems at once, says Chuang. Enterprise-grade chatbots do have their limitations and the pricing is only accessible to large corporations such as financial institutions.
A geek at heart
Another reason Pand.ai is carving a niche in the financial sector is that Chuang, who hails from Johor, once served as head of digital banking at Standard Chartered Bank in China. He has both international and regional experience in the financial sector.
Despite having a background in finance, Chuang is a geek at heart. He completed his master’s degree in engineering at the Massachusetts Institute of Technology and his undergraduate thesis focused on quantum computing.
Chuang’s interest in technology led him to meet David Low, who became the co-founder and chief data scientist at Pand.ai. They met on Linkedin when Chuang was looking for an AI expert and Low was serving the Singaporean government as a data scientist.
“We talked for nine months before deciding to start Pand.ai together. We were bouncing ideas off each other and getting to know one another to see if we could work well together. Now, he heads all of our AI-related research,” says Chuang.
Pand.ai also provides analytical reports to help clients make data-driven decisions. The analytics include basic metrics such as the number, duration and length of conversations held per day, as well as sentiment analysis, which detects the point at which users stop responding to conversations or where they sound angry.
For use cases where clients need to tap into sensitive customer information in their core database, Pand.ai provides on-premise deployment. It installs the AI chatbot on the client’s server instead of the default cloud-based solution to adhere to the Personal Data Protection Act guidelines.
However, much of Pand.ai’s focus is on extending the boundaries of what AI is capable of, such as coming out with its own proprietary technologies. It claims to be the first company in the region to roll out transfer-learning-based natural language processing (NLP) engines, which are the brains behind chatbots. The technology allows it to generate NLP engines with significantly less data input but at the same level of accuracy.
Pand.ai currently supports English, simplified and traditional Chinese, Bahasa Malaysia and Bahasa Indonesia, while Thai is in the pipeline. Adding new languages to the ecosystem is a significant investment, says Chuang.
“It is tough if you want the chatbot to be robust with a higher degree of accuracy. It also depends on what kind of languages we are talking about. In the case of Chinese or Thai, more steps are needed compared with romanised languages such as English or Bahasa, where you have to use a spacebar to denote two different words.
“It boils down to the language structure as well as how much data is available for that particular language. For NLP in general, English and Chinese tend to perform better than other languages because of the sheer volume of data that is available on the internet.”
Chuang is extremely positive on the outlook for AI chatbot technology. Currently, Pand.ai’s chatbot is deployable on most platforms, including Facebook Messenger, WeChat, WhatsApp, clients’ apps and websites. However, he believes that chatbots will eventually become standalone tools for businesses to communicate with their users.
“I think it is inevitable that one day, chatbots will become the ‘new website’. It will be a platform, a new channel for businesses, in its own right. This will start with larger corporate entities before the technology becomes more mature and accessible to the mainstream. I do think AI chatbots will be ubiquitous in the business world,” he says.
In the meantime, Pand.ai has no plans to diversify into other industries in the near future. “For the most part, we will continue to focus on the financial sector so that we can bring the best products and services to clients. But it does not mean that we will not consider other projects. If there are interesting and meaningful projects that we think we can contribute to, we will definitely consider them,” says Chuang.