MANILA, Philippines – Chatbots are not new — you’ve probably encountered one while shopping or when trying to inquire about products and services online.

However, Ragde Falcis, chief executive officer and co-founder of ChatGenie, said that their customer engagement platform works differently. Prospective customers would not have to deal with pressing buttons for an automated response.

He even claims most customers won’t even know they are dealing with AI.

“What differentiates ChatGenie from other AI chatbot platforms that BPO [business process outsourcing] workers can use is that we are using this what we call multi-AI agent framework,” Falcis told Rappler in an interview.

“The way we describe it is technical but to put it in simple terms — imagine you have multiple AI agents that break down a specific customer service inquiry unable to come up with an accurate response.”

ChatGenie was founded in 2018 with the goal to provide an AI-powered chatbot for Facebook Messenger. It then ventured into becoming an in-app commerce platform, much like the WeChat Mini Program.

The platform now focuses on multi-agent customer engagement, backed by investors such as Techstars, Accelerating Asia, and Ideaspace Ventures.

“We signed up two enterprises and we’re in the final stages of discussions with two additional telco enterprises,” Falcis said.

Marco Casalaina, vice president of products at Azure AI of Microsoft, wrote in September 2024 that multi-agent systems are getting popular due to their “ability to provide checks and balances.”

ChatGenie describes it as having several virtual helpers, which then allows BPO employees to focus on other tasks that “require empathy, creativity, or complex decision-making.” Having multiple agents also prevents the bots from “AI hallucination” or when AI starts to give random answers to a prompt. (READ: Language AIs in 2024: Size, guardrails and steps toward AI agents)

ChatGenie’s multi-agent process

When a merchant gets an online inquiry about the price of one of their products, this would first go through an intent agent. The AI-powered intent agent would interpret the inquiry to something that can be understood by AI correctly.

The inquiry will then be turned over to the prompt guard AI agent, which ensures that the question is related to the merchant’s business operations. This means that for a product inquiry, the prompt guard agent checks whether that product actually exists in the inventory of the merchant.

“As you know, we have this new AI trend… there are malicious attempts like prompt injection that can expose data through the AI prompts,” Falcis said.

“That’s one of the roles of a prompt guard agent, to safeguard the data from those malicious steps.”

Once the prompt guard AI agent gives the go signal, the classification AI agent then determines if the inquiry can be considered a general query. The conversation AI agent then drafts a response.

ChatGenie’s AI-generated responses depend on, among others, uploaded information from the merchant — including product information, check out configurations, delivery areas, shipping options, as well as details on payments, discounts, and promotions. The merchant can also provide frequently asked questions for general inquiries.

The final stage of the process is the refinement AI agent, which ensures that the AI-generated response is accurate and relevant.

Falcis said their subscription model for the platform starts at P50,000 per month. Meanwhile, a starter plan, which allows sellers to set up shop on Messenger, Instagram, GCash, and Viber with ChatGenie, starts at P1,000 monthly.

Humans still needed

Falcis acknowledged that the general public still is still lukewarm toward the adoption of AI in work operations, noting: “If they saw that they are conversing with the chatbot, they will just disengage right away.”

At the moment, when you’re chatting with ChatGenie, it does not disclose that the chat box is powered by AI.

However, the platform has another agent — the sentiment analysis AI agent — that alerts human agents onboard if and when a customer inquiry needs to be addressed by a human already.

“If the messages of the customer shows a negative sentiment, then it’s a call to action for the customer service agents that they have to talk to the customer right away. That’s something I think we developed that can help BPO operations only focus on the ones that show negative sentiment,” Falcis said.

Filipino BPO workers see AI as a threat to their job security.

However, the IT and Business Process Association of the Philippines have noted that so far, the impact of AI in business operations have been positive. Although, the group noted it is important for the workforce needs to learn how to adapt the new technology. – Rappler.com