Is CP Group scion’s AI startup worth the $60m it raised?

When Amity Solutions, a Bangkok-based generative AI startup, raised US$60 million last month, it made headlines.

Founded by Korawad Chearavanont, heir to Charoen Pokphand (CP) Group – one of Thailand’s largest conglomerates with US$65 billion in revenue – Amity Solutions is already planning to list on the local stock exchange by 2025.

Amity Solutions CEO Korawad Chearavanont / Photo credit: Amity Solutions

Amity was not always genAI-focused – it started as an employee engagement platform called Eko. In 2020, Eko acquired two more startups, Upstra and ConvoLab, and formed the parent company Amity.

“We’re not trying to build a foundational AI technology that will rival those offered by those in Silicon Valley because we know we’ll never be number one in that,” Chearavanont tells Tech in Asia in a recent interview.

So is this AI wrapper by the CP Group scion worth the US$60 million it raised?

Not the typical genAI chatbot

The company says that Amity Bots, its AI chatbot solution, can understand any user question and provide responses based on the specific needs of an enterprise client, such as customer engagement.

Touchapon Kraisingkorn, Amity’s chief technology officer, says its chatbot only needs one document as reference to answer customers’ questions.

“The problem with [a traditional, non-AI chatbot] is there are a thousand ways to ask a question, and a thousand ways to answer it, so it was difficult to scale,” Kraisingkorn says.

The CTO compares Amity Bots’ data processing to training interns for customer engagement, where they may not follow strict scripts but use manuals for guidance.

Kraisingkorn acknowledges that traditional chatbots are still 10x cheaper than Amity’s AI chatbot. However, he believes that Amity Bots is still more cost-effective for enterprise clients compared to building their own solution.

For instance, Amity Bots doesn’t need to use ChatGPT for every user request. Kraisingkorn says each conversation is “cached” and stored in a smaller model, which the chatbot references for every question.

The familial clientele

Not taking things for granted