servion logo Crafting CX Solutions
subscribe

Get the Servion Blog updates in your inbox.

Your ultimate CX destination

The pressure is on. Your call center is maxed out, you don’t want to add any more telephone agents, your company is going through a digital transformation, growth is not what it used to be, cost reduction is in the air, and everybody is asking you what you can do in terms of automation.

The established vendors you interact with tease their chatbot capabilities, startups keep calling, and you read about ChatGPT (released by OpenAI that has just received a $ 10 billion investment from Microsoft) and Google’s response to it, Bard. You look at analysts' reports and read about alarming failure rates, despite claims that AI is ushering self-service into a new era of customer convenience and response efficiency.

You’re not alone. It’s not a consolation; it is a depiction of the state of the industry.

Let’s unpack this. Undoubtedly, digital solutions have improved over the last year, sometimes dramatically. Speech recognition and natural language processing and understanding (NLP/NLU) are much more capable and open new perspectives for chatbots and customer service. One crucial consideration to remember is that general-purpose tools do not meet the needs of a particular enterprise right off the bat. Tools such as ChatGPT, part of the large language model (LLM) class, need to be thought of in the context of intents (I want to return a product), entities (it’s a Samsung Q70A) and workflows (find the order number, confirm the return, close the transaction), and developing these requires serious work.

For example, ChatGPT doesn’t necessarily know your product catalog well. More critically, it may supply incorrect responses to a customer query; it needs to learn the specifics of the environment in which it will run. To illustrate that point, I asked ChatGPT to list the characteristics of a bike I am interested in. The answer came back with five bullet points, the qualitative ones (good endurance bike for long comfortable rides) were correct, and the hard facts, such as the frame’s material, were incorrect. If ChatGPT were used natively as a chatbot by that bike company, it wouldn’t work well.

The good news is that vendors provide solutions to make that work more accessible and more automated. The not-so-good news is that each vendor toolkit is somehow proprietary, meaning there is no portability of your intents and workflow libraries between vendors.

In a fast-changing landscape, proprietary solutions are fraught with danger. It would behoove you to assess the viability of any chatbot proposal carefully. From an industry observer's standpoint, too many solutions look similar and claim to be the best at one side of the technology or the other. It is a safe bet that many won’t survive, and the industry will coalesce around a few players who excel at developing or using mainstream LLMs. To make your chatbot choice as future-proof as it can be, consider the following:

  • Has the core speech and NLP/NLU technology staying power? Could it be swapped if the company could not sustain the ever-growing investment needed to stay competitive?
  • How are tools such as ChatGPT part of the solution? Several industry leaders have embraced it, making it, if not a standard, a leading solution already.
  • How powerful are the intent discovery, categorization, workflow creation, and management tools?
At Servion, we specialize in helping our customers deploy customer experience solutions, and we deeply understand the technologies mentioned in this post and how they can help transform their customer service. Contact our CX experts for a discussion about your bot strategy.

About the Author

Laurent
Laurent Philonenko

Laurent is the Group CEO of Servion and its group companies. With 30+ years of experience, he served in leadership roles across Avaya, Cisco and Genesys. Laurent finds his zen moments by running and biking. His passion for the culinary arts also keeps him enthused in the kitchen.