Chatbots have come a long way. For years, they were limited to responding with predetermined answers that followed a simple logical structure. But clients can have complex problems, and no tree diagram of possible answers can have enough branches to account for all edge cases that arise. Fortunately, the emergence of large language models has finally made chatbots useful. Armed with mountains of data, startups are now leveraging genetic AI to build custom chatbots for all kinds of businesses and use cases, especially those where people want to be sure about what they’re buying.
of Thailand HD builds chatbots aimed at one such industry: healthcare. The company started as a marketplace for third-party healthcare and surgical services and sees a strong case for developing conversational AI for the healthcare customer journey.
“The products we sell are not the typical things you buy on Amazon. It’s a hospital service, so people shop the same way they do offline,” co-founder Sheji Ho told TechCrunch.
Although every product has a description on HD’s HDmall marketplace, Ho says people still prefer to ask first. “90% of chat messages are people asking for product information. The process of chat trading [is similar to] the offline experience,” he explained.
To further its AI ambitions, HD recently raised a $5.6 million Series A round led by SBI Ven Capital, a subsidiary of Japanese financial giant SBI Group, through its joint fund with Kyobo Securities from South Korea and NTU Singapore’s NTUitive. M Venture Partners, FEBE Ventures, Partech Partners, Ratio Ventures, Orvel Ventures and TA Ventures also participated in the round.
AI for Southeast Asia
Ho says HD is working to build the “Sierra AI of the Southeast Asian healthcare industry.”
Over five years, Ho and his team saw that the faster HD reps responded to questions, the higher the conversion rate. “So there is a very good case for using artificial intelligence to automate this process,” he said. Not only does the company expect conversational AI to help reduce costs, it will allow staff to focus on higher-value tasks, such as answering more complex customer questions.
But Ho and his team seem to have a realistic view of what they can achieve. They won’t be able to match the American companies they have “Nearly unlimited access” to powerful GPUs, talent and venture capital, so the company is focusing on building vertical AI, with local data as its moat.
“Emerging markets need to compete and take advantage of AI using the data they have — proprietary data that no one else has,” Ho said. “We see it happening in other places. Some call this vertical AI, where they use vertical domain data that is proprietary to a particular business or industry. Then they build on top of that and refine the model to the point where they have an AI application that is practical and can start monetizing.”
HDmall. Image: HD
Therefore, HD plans to train the chatbots with the data of anonymous transactions, conversations, FAQs and product catalog that it has accumulated over the years. Currently, 30% to 40% of the company’s transactions are done through chat commerce with customer service workers.
HD plans to use the new capital to launch a chatbot in its market within 3 months and open the technology for use by third parties by the end of this year. Potential customers are hospitals and clinics that need 24/7 customer support. The startup has already partnered with around 2,000 healthcare providers in Asia, which will allow it to improve its basic language model for the healthcare sector. Ultimately, the chatbot service will give the company a new SaaS revenue stream in addition to market commissions.
Fundraising after the pandemic
Like many other startups, HD cut costs and aimed for sustainable growth during the COVID-19 pandemic. The company “didn’t necessarily need to scale up” as it headed toward profitability with 2x year-over-year growth after the pandemic ended, but Ho also saw an opportunity to move faster when others were slowing down.
“You hear people say, ‘You have to raise money when you don’t have to.’ If we upload now then everything else will be cheaper. For example, customer acquisition is cheaper because everyone else stopped advertising during a recession. Talent acquisition too [costs less] because companies unfortunately fire people.”
Globally, startup valuations have fallen in recent years. HD hasn’t escaped that wave, but Ho says he recognized the benefit of accepting a more modest valuation early on.
“I think it’s pointless for companies to worry about valuation at such an early stage. We have seen that in recent years, especially in 2021, when companies started the race with such high valuations,” he said, pointing to the example of Indian health-tech unicorn Prystn, which lost half of its valuation after a period of frantic growth.
“Because they’ve been pumped into such a high valuation, they’ve been forced to grow extremely aggressively, and that’s leading founders and companies to cut corners. You can’t cut corners when you’re in health care and you’re dealing with people’s lives,” Ho said.