Cape Town-based customer service automation specialist Clevva has applied generative artificial intelligence (GenAI) technology to its virtual agent tools, enabling them to hold bi-directional conversations with humans.
Clevva builds ‘digital experts’ or virtual agents and deploys and maintains these for customer-facing businesses.
Co-CEO Ryan Falkenberg highlights the advantages of its GenAI-powered digital agents, currently undergoing testing within the travel industry. These virtual agents differ from standard chatbots by engaging in bi-directional conversations, rather than one-way interactions.
Traditional chatbots follow predetermined paths, waiting for further instructions after providing an initial response. However, genuine interaction involves not only answering questions but also understanding user needs through probing and clarification.
“Most chatbots try to do this using decision tree logic and it quickly gets frustrating for users,” Falkenberg says. GenAI-powered agents mulate human expertise, creating conversations that feel natural while effectively solving problems.
However, GenAI comes with challenges, such as giving inaccurate responses when not guided by specific enough prompts or search phrases.
Clevva has addressed this challenge by enabling its virtual agents to act as ‘prompt engineers’, analysing each customer's request before asking GenAI to generate a response.
Another challenge with Gen AI is that it still can’t be trusted to manage customer interactions unsupervised, Falknberg says. In the contact centre space, Gen AI is being deployed to support agents, but organisations still want human oversight.
Clevva creates constraints around what questions a Gen AI-enabled agent can and can’t answer, pointing to specific datasets it may use and those that are out of scope.
“With a virtual agent they can control the conversation within the company’s accepted boundaries and ensure that every engagement is managed according to clear rules and processes,” Falkenberg says.
The cost factor
Regarding costs, chatbots are typically a lot cheaper than virtual agents as they are a lot easier to set up. But chatbots tend to frustrate customers, who then want to be transferred to human agents. Virtual agents are designed to handle more complex conversations.
“The metric is the cost per conversation,” Falkenberg explains. “Virtual agents cost significantly less per conversation than human agents in many ways. They can handle higher volume and resolve issues quicker than humans, so the total cost per conversation is a fraction of what human agents cost.”