AI Customer Service: The New Era of Support

As a business owner, I’m constantly exploring innovative ways to integrate AI Customer Service into our operations for enhanced productivity. The AI revolution is reshaping our perspective on routine tasks, setting a new standard for efficiency and customer satisfaction.

AI in Customer support has always been a controversial topic since you want to automate the repeating questions to save bucks, and at the same time, you don’t want to push away your customers for a bad customer support experience.

As the CEO of a Customer Support Tool, I’ve started seeing a surge of requests for “AI in Customer Service” toward our Sales and Support reps.

With an average support ticket cost of $15.56, no wonder why you want to automate as much as possible your support requests.

Would you use a Live Chat with ChatGPT to support your SaaS users?

After teaching it with help center articles and past conversations, would you feel confident enough to go live?

— Luca Micheli | Customerly (@iLucaMicheli) January 25, 2023

While I was always skeptical about integrating AI into the Customer Support touchpoints (and 54% of the votes on Twitter expressed the same concerns), we have now found an incredible way to help our customers speed up their workflow with AI.

Here’s how…

While in my previous post on How to NOT use ChatGPT for Customer Service I was still highlighting the lack of personalization and understanding of the inquiries from the AI, today I’ve got incredible news for you.

I was wrong 🙃.

The new layer of Conversational AI we released is fantastically handling customer requests.

Now the AI Assistant is helping our teammates to quickly reply with one click, 90% of the replies to our customers. Like this…

And it is working.

Before understanding how it works, we must cover the basics and some examples. So…

We all saw what ChatGPT is capable of, but how is it doing it, and what is AI?

Artificial Intelligence (AI) combines large amounts of data with fast, iterative processing and intelligent algorithms.

This allows the software to learn (I recommend watching this Ted Talk) automatically from patterns and features in the data. AI is a broad field that includes many theories, methods, and technologies, as well as the following major subfields:

• Machine learning automates analytical model building. It uses methods from neural networks, statistics, operations research, and physics to find hidden insights in data without being explicitly programmed where to look or what to conclude.
• Neural networks are a set of algorithms modeled after the human brain that are designed to recognize patterns.
• Natural Language Processing (NLP) is the ability of a computer program to understand human language as it is spoken. NLP is a key component of Artificial Intelligence.

ChatGPT, developed by OpenAI, is a state-of-the-art Artificial Intelligence model that uses machine learning techniques to generate human-like text.

It’s trained on a diverse range of internet text.

It generates responses to text inputs by predicting what comes next in a conversation.

It’s a powerful tool that can generate creative, engaging, and nuanced text, making it a valuable asset in customer service, content creation, and more.

So how can you train ChatGPT on your company knowledge base in order to let it reply to customer service requests?

That’s the question I’ve got since I tried the first version of GPT and here I’ll share all my findings and how we managed to make it frictionless for people using Customerly.

Our first experiment with AI customer service was pretty basic, I’d say.

The current concept behind it is still so simple tho.

You have a source of truth (your knowledge base) where you share, store, and organize all the articles.

Another great source of information is the canned responses in your Customerly Project. These are the kind of responses you use with your customers, but you don’t share publicly on a knowledge base.

When you have a data source well organized, like our help center here, you will need to clean the content and use it to create an index, and after that, you can find an answer to a customer inquiry using OpenAI APIs.

This first iteration of AI in customer service wasn’t great, and the average CSAT was low due to the lack of context and personalization.

Our latest iteration of AI Customer Service is just incredible.

How are we training ChatGPT to reply to customer support requests?

Before getting to the actual process, we optimized the cleaning process of the source of truth in the best way possible.

So we started eliminating all the unnecessary words and characters from the knowledge base articles and canned responses.

Now that we have cleaned the content, we needed to use an index to store our customers’ help center articles and canned responses.

We decided to use Pinecone for the job and a Belgium pod with Google Cloud to keep things in Europe and GDPR compliant.

Now that the index is ready with all the vectors of our customers that decided to use AI, it can be used by our customers to inquire ChatGPT to gather a great response.

We created Mindbridge.

MindBridge is our internal resource handling all the training requests per customer collections and the prompt towards ChatGPT.

So the complex training process that is happening in seconds when our customer enables the AI in the settings looks like this:

1. Our API fetch from the customer database all the knowledge base articles and canned responses.
2. They are appropriately formatted for the index by our kernel class in the Customerly API (we primarily remove HTML tags and strip the content down to the bare minimum).
3. API contacts MindBridge which will use the content and openai.Embedding to then upsert in Pinecone all the vectors.
4. That’s it.

It seems a very complex process, and indeed it is. This is why we created such a simple solution for our customers.

The only thing you have to do is…🥁

Yep.

You only have to turn it on and choose Formality and Tones.

As I anticipated at the beginning of this post, this AI will massively help customer support agents in many ways.

Let me show you how…

There are different examples of AI examples for customer service, and here are the most common ones:

• Fully AI automated support: the one every business dreams about. This kind of AI would be capable of handling every single support request without interacting with humans.
• AI Suggestions: based on customer requests, the AI will provide a help center article suggestion to read to try to solve the issue.
• AI Assistant: An assistant for customer service reps to provide meaningful messages to share with customers.

Chatbots can identify or predict frictions before a consumer even reaches out for help.

For example, a chatbot can display relevant pages for certain products and services if clicks have trended on specific websites for specific topics in the search engine.

All these chatbot features are included in our live chat solution.

It’s an effective way for a company to provide products to the leads it wants to get.

So if you want to see how AI changes customer service, below is an example.

You would say that this is quite a simple AI implementation, and indeed it is. But the AI Suggestions feature self-served our customer’ customers an astonishing 1,687,884 tickets in 2022.

Now before getting to the Fully AI automated support, you might be wondering if…

Yes and no. It depends a lot on the kind of customer support requests the representatives are handling and the complexity of the support.

Yes, AI will probably replace most support reps handling basic tasks.

But No, AI is still not capable of replacing customer service completely. Why?

Cause even if you manage to solve 90% of the support requests with fully automated AI, 10% must be analyzed and processed by humans.

Also, humans need to overview what AI is doing and how customer satisfaction improves based on interactions.

This is precisely why we first wanted to help the support reps with the AI Assistant.

While AI Suggestions are a great advantage for the entire team to quickly resolve basic customer issues and faqs, the AI Assistant is providing a precise, in-depth, and polite response to the support rep.

Customerly AI Reply for the Perfect Answer Every Time

Customerly AI automatically finds the perfect answer to customer questions and customizes them. This ensures your customers receive accurate and personalized responses, improving their overall experience.

This will change massively the performance of the customer support rep.

In fact, customer support reps that usually underperformed, now with the benefit of AI Assistant, are overperforming the previous leaders.

Some customer service reps are closing 90% more than the number of conversations they were closing before the AI Assistant.

This is how AI is shaping the customer service industry.

But this is only one benefit of using AI for Customer Support.

AI Assistant also Makes Personalized Responses Easy.

All you need to do is write a short custom response, and our AI Assistant will do the rest.

Sometimes the response must be personalized with context about the customer status, data, order, purchase, you name it…

This is why the Write for Me you see below is a great example of how AI is helping customer service agents and creates less friction and speed.

Busy Customer Service Teams Can Now Turn Support Exchanges into Articles

Another way AI is changing customer service is by turning support exchanges into articles has become a 1-click solution.

Experience the ease of transforming customer support interactions into ready-to-publish help center articles with no extra effort on your team.

The cool part about this is that once the article has been published, they will train the AI in seconds, and from now on, you will be able to use the AI reply that is fully capable of handling the same request the next time.

If you are used to writing from scratch your FAQs articles, this is a great revolution the Customerly AI assistant is bringing to support managers or reps.

AI Assistant Summarization

Last but not least, with the help of the AI Assistant Summarization, the AI is helping your customer support team to summarize the conversation quickly to share it with the tech team or someone else that should take actionable insights.

Now instead of reading through the whole conversation, your colleagues can enjoy highlights and actionable.

This is another way AI is shaping the customer service world. As in any other industry, AI is also speeding up workflows for customer service.

A simple chatbot might be the most common customer support tool or the one that the average consumer might encounter frequently.

Chat bots can be effective in removing agents’ needs and reducing their workload.

Chatbots can provide canned answers or search existing sources like help center articles, web pages, or other interaction history. If the query includes half of the customer service tickets of a company, this will save the agent time.

While Chatbots must be programmed with some paths, the Customer Service AI tools will not.

Usually, a chatbot must be programmed by customer support managers with the choices you want the customer to follow, and based on the choice the bot will reply or provide the right agent.

Most tools like the ones you saw in this article work out of the box without bothering the customer service teams to create such flows.

This is why we are working on a more powerful chatbot, including Chatflows and AI.

The challenge to provide fully automated support will be giving the customer service teams flexibility and integrations with other tools.

This is why we believe that Chatflows, in collaboration with the AI Assistant we have created, will bring up to 80% of the number of support requests that can be handled automatically.

In a world with an international customer base, it becomes essential that business offers multi-lingual support.

There might not even be a single seller who can offer support for all major languages, but AI can support this.

Whether the requests are a mix of Turkish, Italian, or English, AI will understand the context of your messages and provide a quick response.

The AI Assistant will always provide the response based on the customer support rep’s browser language (because it is the one that can read) and then the Customerly inline translation feature will translate the final message.

AI can help customers improve service by automating the handling of simple problems faster and reducing costs for the user and businesses.

It’s also a competitive advantage for businesses since the rapid adoption of AI from a competitor of such AI tools might outperform your support team and generate better customer satisfaction.

You need to adapt as quickly as possible a solution that will help you speed up and optimize your customer support workload without sacrificing customer satisfaction.

The application can make customer service easier by optimizing the customer experience and providing them with more resources for solving problems.

There is a catch tho. You don’t want to push away your customers because the AI replied with something totally useless.

I’ll give you an example.

I’ve been trying to reach out to the Google Ads support team, and the only responses I’ve got are from AI.

You can tell because it (I still don’t know which is the correct pronoun for the AI) is sending me the same responses without reading the context of my replies.

This creates frustration on the customer side, and you don’t want to degrade your customer experience.

So my suggestion to transition to AI automated support is to:

• As highlighted above, train the model with your knowledge base and canned responses.
• Test out the responses yourself and see the quality of them
• Once you feel confident that you are trained to answer your customer questions, you can create a chat flow to provide an answer from AI directly.
• Monitor with a CSAT survey and see how AI affects your overall customer satisfaction.

Getting started with AI for customer service is a transformative step towards supercharging your support workflow and boosting your team’s efficiency.

It is easy to start implementing customer service automation if you use the correct tools.

First, you will need to have your Knowledge base or Helpdesk software on a provider that is integrated with AI.

Thankfully for you, with Customerly’s Conversational AI, you can leverage a suite of AI-powered features designed to streamline your customer service operations.

In fact, you can use the Help Center + Conversational AI + Live chat as 3 great companions to quickly adopt the most advanced AI for customer support.

The AI Suggestions feature provides your customers with instant, accurate responses to their inquiries right within the chat interface, significantly speeding up customer queries.

The AI Reply feature automatically finds and customizes the perfect answer to customer questions, ensuring your customers receive accurate and personalized responses, thereby improving their overall experience.

The ‘Write for Me’ feature simplifies crafting personalized responses. You only need to write a short custom response, and the AI will do the rest.

The AI Summarize feature turns complex customer support chats into short, easy-to-understand summaries, making it easier for your team to manage and respond to customer queries.

Moreover, you can effortlessly transform customer support interactions into ready-to-publish help center articles, reducing the time and effort required to create helpful resources for your customers.

By signing up for a 14-day free trial on Customerly, you can experience firsthand how this AI tool can elevate customer service to the next level.

Don’t miss out on this opportunity to revolutionize your customer support and give your business the competitive edge it needs.