Article | 1 min read

ChatGPT for customer service: Capabilities and limitations

Learn about the capabilities and limitations of ChatGPT for customer service and how businesses can leverage OpenAI's API.

By Hannah Wren, Staff Writer

Last updated October 26, 2023

ChatGPT is an impressive conversational AI (artificial intelligence) chatbot by OpenAI that launched in 2022. Since its release, users worldwide have been using the tool to write, code, and find answers to general questions. But can OpenAI API or ChatGPT be used for customer service?

In this article, we’ll answer that question, explain the capabilities of OpenAI’s large language model (LLM), and provide best practices for using the OpenAI API for customer service.

What is ChatGPT?

ChatGPT (Chat Generative Pre-Trained Transformer) is an AI chatbot with a conversational user interface, enabling the tool to mimic human communication.

Cognitive technology, such as artificial intelligence (AI), natural language understanding (NLU), machine learning (ML), and natural language processing (NLP), train the bot to understand context and human language patterns. It can then reply to inputs with human-like dialogue. See additional information about this technology in these in-depth guides:

ChatGPT overview

Release date: November 30, 2022
Developer: OpenAI
Engines: GPT-3.5 and GPT-4
Knowledge cutoff: September 2021 for GPT-3.5 and
no cutoff for GPT-4 (with a third-party plug-in)

Upon ChatGPT’s release, the bot was primarily a tool for general research, education, and creativity. Now, businesses are finding ways to use the bot for internal processes to help their employees work more efficiently.

Customer-first AI service providers like Zendesk use OpenAI’s API to enhance generative AI features and help agents streamline internal tasks. By leveraging solutions that use OpenAI’s LLM, businesses can bring generative AI to the forefront of agent tools for helping with content summarization, building a knowledge base and macros (pre written responses), expanding agent responses, and more.

Can businesses use ChatGPT for customer service?

ChatGPT wasn’t built as a customer service tool and isn’t ready to interact with customers—at least for now. As a result, we would not recommend using ChatGPT for customer support at this time.

However, many customer support teams are using ChatGPT internally. Here are a few examples of how they can help agents as well as customer service leaders be more productive:

  • Edit and translate help center articles.
  • Suggest customer service interview questions for the hiring process.
  • Help write questions for customer surveys.
  • Polish team meeting notes.
  • Provide insight for agents responding to customer reviews.
  • Recap Zoom calls with customers through transcript and action item summaries.

Note: Privacy concerns should be top of mind, especially when using ChatGPT solutions for customer interactions.

What are some limitations of using ChatGPT for customer support?

ChatGPT does have several limitations for customer service. Many of these are key reasons businesses should only use the chatbot internally for now.

For instance, the chatbot sometimes:

  • “Hallucinates,” or makes up fictitious information and presents it as fact
  • Makes logic and reasoning errors
  • Poses potential security issues when fed customer information
  • Is unable to answer highly specific or niche questions
  • Offers unintentionally biased or discriminatory responses that may harm your brand’s reputation

Though ChatGPT has a knowledge base cutoff of September 2021, there are plug-ins you can use to crawl the web. However, these plug-ins are only available through the paid version and integrate with third parties. ChatGPT has a web crawler on the horizon, GPTbot, but websites can choose to block it.

This may be minor in some industries, but ChatGPT may be unable to keep up with business and customer needs in rapidly evolving fields like tech and medicine.

How OpenAI’s API can improve customer support

Although ChatGPT has made outstanding strides in generative AI, it isn’t a good fit for customer service on its own. Fortunately, businesses can still leverage the powers of OpenAI’s large language models by looking for a customer service solution that integrates with the OpenAI API.

OpenAI API graphic

Before we dive in, let’s first answer an important question: What’s an API? An application programming interface (API) is a set of programming protocols that helps two or more programs talk with each other. In short: An API is responsible for how your business shares information and streamlines your processes.

According to the Zendesk Customer Experience Trends Report 2023, business leaders say that improving agent workflows is the fifth-most popular use of AI in their workplace. With a superpowered customer service software that uses OpenAI’s API at your disposal, you can enhance your workflows in the following ways.

Deliver faster support to customers

Businesses with an automation-first strategy can use OpenAI’s API with their customer service software to deliver faster human support to their customers through ticket summarization. For example, if a customer submits a long message about a product issue, the agent would receive a summary of its main points to speed up average resolution time.

Beyond the capabilities that OpenAI itself enables, using a customer service solution like Zendesk can unlock even more AI capabilities for faster responses, including tasks like:

  • Revealing the customer’s intent and sentiment
  • Directing agents toward possible solutions

As mentioned, this enhanced customer support experience will significantly improve your time-to-ticket resolution.

Improve agent productivity

Your customers will know when they receive a run-of-the-mill reply, which may negatively affect their CX. People often associate automation with canned responses, but support agents can leverage the generative AI tools to enhance the quality of their replies and customer-facing ticket comments without spending too much time writing a custom reply for every customer.

Agents can do this by asking the software to:

  • Expand replies by adding conversational context and creating a customer-ready response. For example, if an agent wants to say “One sec,” AI can expand the reply to say “Please give me a moment as I look for a solution.”
  • Change the tone to be more friendly or formal, depending on the customer.

Elevate knowledge management

With OpenAI’s API, businesses can level up their help center and knowledge management capabilities. For example, the API builds on the proprietary system of Zendesk to scan inbound support requests for trending and common customer problems for missing support articles. If support teams notice a specific customer problem is trending, they can leverage the Zendesk and OpenAI API integration to instantly create and edit a help center article. Chatbots can then leverage it for future customer queries.

Transform the customer experience with AI

Bid adieu to clunky, outdated support processes and ring in an era of empathetic, effortless AI-powered support. Read our free report to uncover tactics for leveraging AI to achieve successful CX outcomes.

Frequently asked questions

Below, we answer the most common questions about using ChatGPT for customer service.

What are large language models?

A large language model is an AI system that uses deep learning and copious amounts of data to imitate human intelligence. LLMs allow chatbots to accurately comprehend and reply to prompts.

Large language graphic

What is the future of ChatGPT and customer service?

Though ChatGPT is adept at enhancing internal processes for customer service agents, it’s unlikely that it will replace human agents completely. Human empathy and understanding, particularly in delicate or complex situations, will likely remain crucial. However, a combination of AI and human customer service is likely to become the norm, with AI handling simpler queries and humans stepping in when necessary.

What are the capabilities of ChatGPT?

ChatGPT is a vast AI chatbot with several capabilities, including:

  • Creating content
  • Editing for grammar and syntax
  • Writing code
  • Helping with brainstorming
  • Simplifying complicated concepts
  • Carrying conversational context from one prompt to the next
  • Translating other languages

What are the different ChatGPT models?

OpenAI has developed several versions of the ChatGPT model over time, each with improvements and refinements. The following are some of the different versions:

GPT-1: The first in the series, GPT-1 had 117 million parameters. OpenAI first trained it in June 2018.

GPT-2: Released in February 2019, GPT-2 was a substantial improvement with 1.5 billion parameters. It was not initially released in its full form due to concerns about potential misuse.

GPT-3: Released in June 2020, GPT-3 took another significant leap with 175 billion parameters, making it one of the most powerful language models to date.

GPT-3.5-Turbo: It’s similar in capability to GPT-3, but it’s fine-tuned further to provide a better user experience in single-turn tasks and multi-turn conversations.

Meanwhile, GPT-4 is the premium version of ChatGPT. It can do everything that 3.5 can do, as well as:

  • Allow its users to create tools
  • Offer a longer memory of 64,000 words
  • Pull the most recent data from the Internet
  • Understand complex inputs
  • Accept image prompts

Remember that each subsequent version is essentially a continuation of the same approach. They are trained on a broad range of internet text but don’t know specifics about which documents were in their training set or have access to any private or proprietary information.

OpenAI and Zendesk

Through the Zendesk collaboration with OpenAI, we use generative AI capabilities coupled with decades of CX data to improve the customer experience. Though Zendesk already offered an out-of-the-box CX solution, our existing capabilities and ability to scale are only enhanced by further enhanced AI-powered features, like:

  • Macro creation
  • Expanding replies
  • Summarizing content
  • Knowledge base management

Transform the customer experience with AI

Bid adieu to clunky, outdated support processes and ring in an era of empathetic, effortless AI-powered support. Read our free report to uncover tactics for leveraging AI to achieve successful CX outcomes.

Transform the customer experience with AI

Bid adieu to clunky, outdated support processes and ring in an era of empathetic, effortless AI-powered support. Read our free report to uncover tactics for leveraging AI to achieve successful CX outcomes.

Get the free report