OpenAI’s ChatGPT has taken the world by storm. Its ability to answer questions and generate high quality text on-demand is astounding. At Y Meadows, our mission is to use AI and robotic process automation to empower customer service departments to provide better, faster, and easier service. Due to ChatGPT's impressive capabilities, our clients have asked us if we can help them utilize it to automate customer service responses. Our answer is: Yes! We’re pleased to announce that you can now use Y Meadows to integrate OpenAI with your knowledge base and ticketing system.

Take a look at what you can do with Y Meadows and OpenAI:

In this example, we have an airline company called Underground Airlines. They use Front as their shared inbox for customer support. A customer named Anna is flying with them next week but she's 38 weeks pregnant, so she sends a message to their support team asking if it's okay for her to fly.

Once the email arrives in the company's email/ticketing system, a Y Meadows Journey is triggered.

In the journey, Y Meadows will take the client’s email and search the knowledge base for relevant information. OpenAI will write a response and add it as a ‘draft reply’ in Front. Then, a human can check that the email is correct and send it out! This not only improves resolution times, but it also reduces the effort needed by support agents to accurately find information and respond accordingly.

After Y Meadows searches the Knowledge Base for relevant information, this article appears as its top result:

The information found from the Knowledge Base search step is then used by Y Meadows integration with OpenAI to draft a reply:

Afterwards, the agent can just click 'Send' and they're done!

With the help of Y Meadows and OpenAI, customer service teams can work through tickets faster and handle larger volumes of requests since the AI does the research and the writing. This allows agents to apply their focus on higher value, more complex requests while also providing customers with answers faster than ever before.

If you’re interested in learning how the technology works, please read the next section. If you would like to learn more about how to apply this technology in your business, please reach out to us at info@ymeadows.com or learn more at www.ymeadows.com.

How It Works

OpenAI has created a family of AI models called GPT-4 (Generative Pre-trained Transformer 3) and offers an API which allows other systems to integrate and use its models. The GPT-3 model only knows the data it was trained on; it can’t answer questions about information that was not in its training set. For customer service, most answers are based on a company’s policies or knowledge base. We can add that information to make the AI model aware of what it needs to know.

GPT-4 can also be used to search text. The way this works is that GPT-4 can convert words to vectors, a series of numbers that represent the meaning of the words. We can take a search query and convert it into vectors. We can also convert the knowledge documents that we are searching into vectors as well. Then, we can can calculate which KB article vectors are mathematically closest to the query’s vector. The advantage of doing this is that it can find matches even if the word you are searching for is not in the KB article. For example, “problem logging in” might turn up a page that uses the phrase “authentication issue”.

One of the main issues with large language models is that they generate sentences based on statistics of how words are used on the Internet. It’s not really “thinking”, and it can sometimes give wrong information, yet still look correct. This can be mitigated by having a human review what the AI is going to send out. In the example above, we used a draft email message to do that. Another option is to use the Y Meadows review module. The review module allows support agents to quickly see the original message and the proposed response. They can edit, approve, or reject the proposed response.

The Y Meadows integration with OpenAI has other uses in addition to generating responses to messages. 

For example:

• We can ask the model to rewrite an email written by a non-native English speaker to correct spelling, grammar, and/or to adjust the tone of the message.
• We can ask it to summarize a message.

OpenAI is a great tool, but it’s just an API. To integrate it into your customer service environment you need to be able to trigger it on every ticket, gather data from internal and external systems, template prompts, and integrate with email and ticketing systems. Y Meadows is a customer service automation platform that can do all of that and more!

The example and screenshots above use Front. But, Y Meadows integrates with ZenDesk, FreshDesk, Gmail, Jira, Helpscout, Microsoft Office 365, Salesforce Service Cloud, and many more!

Interested in getting started with Y Meadows AI customer service automation solution?  Contact us at info@ymeadows.com or learn more at www.ymeadows.com.