Is ChatGPT suitable for Customer Service?

Leave a Comment

Is ChatGPT suitable for Customer Service?

ChatGPT from OpenAI created a lot of buzz about its capacity to handle multi-turn open dialogues. It has been financed amongst others, AWS, Microsoft and individual investors, and is deemed as a serious competitor to Google’s search engine, about to be integrated into Bing.

Many of our customers ask Expertflow whether ChatGPT can be used for customer service. Expertflow CX is AI-agnostic and enables companies to integrate AI engines both for speech and for chat, such as chatbots, speech recognition, Natural Language Understanding, Voice and Face biometrics or toxic language detection to customer service solutions. AI engines can be used for self-service (such as chatbots or conversational IVR) or for agent assistance to provide answers and suggested actions to the agent. These AI engines include Amazon Lex, Rasa, Microsoft Cognitive Services, perspective API and others.

So the short answer is – yes, ChatGPT can be integrated with Expertflow’s Hybridchat solution, including embedding into a Contact Center through the open Chatbot API. Expertflow would then connect to any chat channel (Whatsapp, Viber, SMS, FB Messenger,…), Social media. It could also use ChatGPT as a conversation engine when combined with a speech recognition engine, to act as a conversational IVR or agent-assist solution.

Do we recommend companies to do so for customer service? No, for the time being, for the following reasons:

  1. ChatGPT knows only the public internet. It has no access to company- internal data and therefore doesn’t know a company’s products and services, customers, and prices. Customer service is all about these. As long as ChatGPT doesn’t provide data ingestion from company-internal knowledge bases such as with Amazon Kendra, Microsoft Cognitive Search and others, ChatGPT will not be able to answer any of these questions.
  2. Customers typically have precise use cases, such as price requests, ordering, updates on a ticket, and changes/ updates on an order. This means that the activities that a customer executes typically require a back-office integration to a company’s ERP, CRM, ticketing, online shop or other line-of-business solution. Access to these internal sensitive API’s available in a controlled way resulting in a coherent customer experience.
  3. Customers require correct answers or actions to their tickets, problems, and order. This can’t be an answer or action that sounds correct – it has to be correct and reliable, so they must initially be curated by human beings. Firm answers or actions that sound correct but are factually wrong would undermine customer confidence.

Some companies, such as AirAsia, have completely dropped customer service in favour of AI. AI has a strong role to play in reducing costs and increasing the availability and reach of customer service. However, Expertflow thinks that the human element will never go away completely. There is always a certain portion of queries that require emotional elements, real-life experience or access to company-internal, access to experts or having a certain cultural background. Expertflow’s portfolio has therefore been to embrace AI technologies in a way that allows seamless hybrid blending of AI and humans in conversations. This approach allows companies to adopt AI technologies gradually without disruptive forklift changes.

Leave a Reply