The evolving role of GenAI in enhancing chatbot automation.

By Craig Smith, CX Automation Manager and Damian Copeland, COO

While artificial intelligence (AI) has been around for nearly 50 years, adoption rates are accelerating across every business sector as the hype around generative AI (GenAI) spurs investment.

In the contact centre space, many operators have already implemented AI technology in some guise to support better engagement and enhance customer experience (CX).

“The dawn of GenAI sparked renewed interest in how operators can leverage the technology to gain a competitive advantage,” affirms Craig Smith, CX Automation Manager at Connect.

Smith explains that the launch of multimodal large language model (LLM) AI assistants like OpenAI’s ChatGPT sparked this industry interest, and affirms the potential these tools have to revolutionise the engagement game.

While the potential gains from AI in terms of cost savings and improved performance are massive, Smith says it is critical to apply the most relevant AI solution to the identified challenge.

“Despite the hype around the application of GenAI in the contact centre, the reality is that operators need a compelling use case to warrant the investment to implement this advanced technology. Rushing to embrace this version of AI in the contact centre can prove costly and may not yield the desired results.”

For this reason, Connect uses a consultative process to identify the main challenges a client faces to determine the most applicable solution, rather than pushing GenAI as a cure-all.

“Due to the hype in the business world, many startups have popped up selling GenAI-based solutions, which often leads to horrific outcomes,” cautions Damian Copeland, COO at Connect.

“Our business analysts go into every exploratory meeting without an AI agenda. They aim to identify and understand the major operational and customer pain points by gathering as much data as possible.”

These data-driven insights provide a holistic understanding of the business needs, which enables the team to recommend the most appropriate solutions.

This is a particularly beneficial approach as discussions around the role of chatbots in the contact centre intensify amid the AI hype, especially as the technology can expand the role and impact these solutions have in the business.

“As a significant number of customers call the contact centre for the same issues, automating specific responses can offload a large volume of calls from live agents to automated chatbots,” says Smith.

The first point of departure in this discussion should determine whether a chatbot should leverage direct dialogue or a more advanced AI interface.

“At a basic level, conducting a thematic analysis can identify key phrases that customers use most often when contacting the contact centre,” elaborates Copeland.

“Assigning intent and outcomes to these phrases can deliver automated direct dialogue capabilities that can support basic transactional engagements.”

Contact centres that require more advanced chatbot automation can leverage conversational AI capabilities to effectively handle the majority of the more complex automated engagements that contact centres currently field, before taking the leap to GenAI-enabled tools.

Traditionally, conversational AI models are built with open natural language processing (NLP) interfaces, designed to allow users to interact with the chatbot using natural language, rather than requiring specific commands or keywords. These capabilities enable chatbots to handle a wide range of interactions in an intuitive and user-friendly manner to enhance overall CX.

“These deployments make up the majority of the AI-based solutions we currently implement in the contact centre environment,” affirms Smith.

“However, the ultimate success of a conversational AI system comes down to the design and detail of the solution, as non-human curated systems are prone to costly errors.” continues Copeland. And this is where many implementations can fail to deliver an adequate return on investment (RoI).

“Businesses that choose to develop conversational AI models in-house must understand the potential pitfalls. When partnering with a service provider, it is critical to find one that has experience in designing solutions, with access to the relevant technologies and best tools to deliver the desired outcome.”

However, that’s not to say GenAI doesn’t already have a role in the contact centre, with various use cases emerging.

“A powerful use case includes real-time agent assist roles, where the GenAI engine listens and summarises the conversation,” continues Copeland.

These advanced solutions allow GenAI to work in the background, feeding relevant info to live agents to support hyper-personalised and relevant interactions in real time.

“This is the safest use-case at present as it keeps humans in the loop because GenAI hallucinations are still an issue,” adds Smith. GenAI is also adept at simplifying technical content for consumer consumption.

However, before the sector can start putting GenAI-enabled chatbots directly in front of customers, operators must consider the quality of the data and information and the knowledge base used to train and augment the AI engine.

“Using the internet to train the engine seldom delivers the outcomes a business wants when their customer calls them, which damages consumer confidence in the technology” states Copeland.

“A better option currently is using a retrieval-augmented generation (RAG) architectural pattern in GenAI design. This approach can help to bridge the gap in solution design as it builds the LLM solution to address specific questions about your business, enhancing the accuracy and relevance of responses,” says Smith.

“When fed quality, company-specific data and appropriate guardrails are implemented, GenAI can also start to handle the speech and conversation interfaces that enhance dialogue between customers and conversational AI chatbots by generating new content that people can understand while removing the potential for human error in the process,” concludes Smith.