Chatbot
istock

Is ChatGPT’s customer service potential worthy of the hype?

by

ChatGPT has been all over the news since its prototype launch in November, as excitement has risen about its potential in many different applications. One of these is in customer service. Could it really live up to the hype? MyCustomer spoke to Gartner’s Uma Challa about the AI breakthrough that everyone’s talking about.

12th Jan 2023

I don’t think it would be unfair to say that the majority of consumers have had fairly mixed customer service experiences with AI-powered chatbots.

Yet despite their hit-and-miss nature, businesses are doubling down on bots, largely thanks to their 24/7 availability and immediate responses, and their cost-effectiveness.

It also helps that - after a rough start - the technology has continued to improve. And with the launch of OpenAI's ChatGPT, the technology may have just taken another massive leap forward. 

So with a history that can be traced back to the 1960s and a growing customer service prominence that has seen interactions expected to exceed 10 billion by 2027, why is this latest chatbot technology causing such a stir?

Chatting about ChatGPT

Launched in November of last year, ChatGPT is, like any other chatbot, an AI bot that provides a conversational interface that allows customers to interact in a natural language.

In layman’s terms, and perhaps because I’ve been watching too much Master Chef of late and Greg Wallace likes to use this phrase to describe an enhanced/more sophisticated version of a classic dish, ChatGPT is a chatbot that’s been to the gym – or perhaps a chatbot that’s been to college would be a more accurate description.

Developed by OpenAI, ChatGPT combines advancements in AI technologies with fine-tuned language models to generate human-like responses in text, making it one of the largest and most powerful language processing AI models in the world.

According to Gartner senior director analyst Uma Challa, these improvements make ChatGPT a completely different animal to your run-of-the-mill chatbot: “The result is a human-like chatbot that is superior in intent understanding, can engage and answer questions of the users, answer follow-up questions, admit its mistakes, challenge incorrect premises, generate contextualised content, translate languages, extract, and organise information, generate executable code and importantly moderate its content.

“The relatively high levels of intelligence demonstrated by ChatGPT compared to any chatbots in the market is the reason for the hype and is triggering the ideas around potential applications of this technology.”

But can this shiny new model undo years of below-par chatbot customer service?

Rehabilitating the reputation of chatbots

Chatbots have historically underwhelmed consumers, and whilst they have undoubtedly improved in recent years, they still continue to be one of the most maligned customer service channels.

Indeed, as recently as 2019, 54% of online US consumers were stating that interactions with customer service chatbots had a negative impact on the quality of their lives.

Arguably, part of the reason that chatbots so often incur the ire of users is that they aren’t being used properly. When correctly implemented, chatbots can be of great value to organisations. They can improve self-service for a broad set of customer queries, allowing human agents to focus on more complex issues for the customer.

Current chatbots operate within set parameters and are not designed to deal with more complex problems.

They operate within set parameters and are not designed to deal with more complex and intricate problems, as they do not allow for intelligent interaction that can improve customer experience on a broad range of topics.

Until now.

Challa believes that ChatGPT will be able to enhance the chatbot experience and provide a far more sophisticated service: “ChatGPT is based on GPT version 3.5, an AI model with 175 billion parameters. The combination of this fine-tuned LLM and chatbot technology, promises significant improvements in chatbot performance by customising the model that can engage in human-like interaction when responding to a question from the user, answer follow up questions and admit to its mistakes.

“It can generate responses in natural language, easily understood by the customers and remember the context of the conversation including the prompts provided by the user and respond in the context of the prompt throughout the conversation. It is also able to moderate its content by detecting if the text generated may be sensitive or unsafe.”

Picking your spot

Whilst ChatGPT may be able to improve the user experience, another of the historic issues with chatbots has been precisely where they are deployed within the customer journey.

Challa believes that this starts to become a problem when organisations begin their chatbot deployment by selecting a technology, rather than a use – leading to an ineffective solution and lower-than-expected improvements in CX and ROI.

If organisations want to increase the likelihood of a successful chatbot deployment they must choose the best-fit use cases before selecting the technologies.

If organisations want to increase the likelihood of a successful chatbot deployment they must choose the best-fit use cases before selecting the technologies.

“While ChatGPT can improve the customer experience by interacting in human-like fashion, organisations should carefully consider the use cases that are best fit for initial and subsequent phases of adoption.

“Initially good use case candidates to begin ChatGPT adoption can be for the following customer intents: FAQs, product information, knowledge articles, self-service guidance and low complexity customer intents that can be resolved with content. As models improve in the level of maturity, complex tasks can be delegated to chatbots in subsequent phases of adoption.”

So, all that’s needed is an upgrade to ChatGPT and a more shrewd deployment system and it’s plain sailing? Not quite.

Viva la revolution

There is no denying that the improvements that ChatGPT is going to make to the chatbot sphere could be revolutionary.

If it delivers on its potential, the shift from chatbots that are only able to deal with very set queries within specific parameters, to those that can engage in human-like interaction and field questions outside of the initial remit will completely transform the customer service industry.

Challa believes that ChatGPT has raised the bar exponentially for chatbots, and if the AI system is applied correctly it can be a gamechanger.

ChatGPT has raised the bar exponentially for chatbots, and if it is applied correctly it can be a gamechanger.

However, he also has some reservations – arguing that the following concerns will need to be addressed before widespread adoption:

  • Accuracy: Models are as good as they are trained and with large models that involve billions of parameters there is a high probability of generating incorrect responses.
  • Influence: The models are learning continuously and bad actors can train the models to deliver responses that can be potentially damaging to the businesses and customers.
  • Verbosity: ChatGPT generates contextual text when answering questions from the user, which can be very verbose at times and may not be succinctly providing the answer.
  • Model as a service: Given the high cost of training and running the model, it is only possible for a handful of companies with deep pockets such as OpenAI to host and operate these models. As other companies adopt the base models as a service to customise, the ownership of the resultant enhanced model, depending on where it lies, may create a significant imbalance.
  • Intellectual property: The model is trained on vast amounts of data from the internet, and it is unclear as to what the legal precedent may be for reuse of this information.
  • Information security: To customise, companies will need to provide information to the base models. How the base models will use the data should be reviewed and understood to ensure the security of the company's information.

Despite these issues, it is undeniable that ChatGPT has the potential to provide far more upside than downside – the journey is definitely worth the destination. 

As Challa puts it: “There are concerns but the potential of the technology cannot be ignored, and organisations must invest to understand the technology, related concerns, and opportunities to adopt where applicable to stay competitive.”

Replies (0)

Please login or register to join the discussion.

There are currently no replies, be the first to post a reply.