AI and Chatbots in Customer Service

Chatbots and AI are quickly moving from the category of trendy, new customer service tools to the mainstream. More and more customer service organizations are turning to the power of AI and chatbots to provide efficient streamlined service to their customers.

Even if you’ve implemented chatbots in the past, new advancements in software have opened up new ways to engage your customers.

What is Artificial Intelligence?

Artificial Intelligence is defined as a computer system that simulates a human’s ability to understand and learn. Before AI, computers needed to be programmed with exactly what they were supposed to do. If this happens, then do this. But with the advancements in AI, humans can tell computers what the goal is, and the AI will learn and optimize a way to get there using algorithms and calculations that simulate the way a human thinks - but much faster.

What is a chatbot?

The definition of a chatbot overlaps with AI, but they are not the same thing. Chatbots are a type of messaging software that interacts with customers and website visitors to gather information and provide help. The most basic chatbots in support use simple if/then statements and are programmed to recognize phrases and respond accordingly. This isn’t AI, but it is automation. More advanced chatbots can use AI to learn and improve their ability to understand what’s being asked of them. If your chatbot only recognizes a set number of keywords, it doesn’t use AI.

Chatbots vs Artificial Intelligence

Chatbots and Artificial Intelligence are two topics that are often lumped together. In fact, we’re doing it right now. But what’s the difference between the two?

Not all chatbots use AI. Not all AI has a messaging interface. But when you combine the two of them together, you get a really helpful AI assistant, or “bot.”

Chatbots are strictly customer facing and they may use AI to better understand customers or to surface better information. For example, the Freshdesk bot called Freddy uses machine learning to “read” existing knowledge base articles and match them with what it thinks customers are asking. The more conversations that Freddy has “read” or learned, the more accurate it will be.

  • Rule-based chatbots: Chatbot software tools without AI functionality are called “rule-based chatbots” because they simply follow the rules they’ve been given.
  • Machine-learning based chatbots: software that learns from customer interactions can infer context, refer to older conversations and offer resources to help answer questions. Because it uses a type of AI called “machine learning” it actually improves over time.

AI is all about understanding - whether it’s being able to read text, detect patterns or recognize an image. This can be really helpful when working with chatbots, but customer service can also benefit from using AI in other ways. AI can provide helpful information on the agent side of a helpdesk or uncover insights based on customer conversations and ratings. Basically, anytime you need a lot of data crunched or processed, AI is going to be a very big help.

Why use chatbots?

Chatbots arose from a very simple request from customers - we want faster answers. With growing companies struggling to meet the demands of incoming questions, they turned to a technology that was often seen in science fiction movies: the helpful robot assistant.

And at first, the technology wasn’t ready for the big stage. Facebook Messenger bots for customer service reported a 70% fail rate when they launched. But when used for a narrow purpose and backed by powerful AI technology, chatbots can actually help provide a range of benefits for customers and for customer service teams.

Instant replies: Forrester has reported that 73% of customers say that valuing their time is the most important thing a company can do to retain them. Customers absolutely love being able to help themselves, find answers quickly and get back to the rest of their lives.

Reduce incoming volume: Every customer service team wishes they had more time in the day so they could spend more time on the things that really matter. Using chatbots to answer simple questions frees up your agents for the sticky, difficult questions. It can also help you reduce costs as your customer base grows. As you get more customers, your chatbots can handle more questions meaning you don’t need to hire as many agents to support the same number of customers.

Chatbots for Customer Service

While they may have gotten a bad rep in the past, chatbots can be extremely helpful for customer service teams, especially in high volume situations where the same questions come up frequently. Chatbots can also make customer service more efficient by gathering information, verifying account data and triaging before connecting customers to help. Let’s look at each of these use cases in detail.

Using Chatbots for Gathering Information

When a customer initiates a conversation, there are a lot of formalities to go through before help is provided. You might need to understand what account they are talking about, then verify that they have the authority to talk about that account using secret phrases and then you need information about their question. This can be a long process, especially if the customer needs to go looking for information. Using a chatbot to gather this preliminary information before connecting the customer to a human can shorten the wait times for customers and make customer support agents more efficient.

Using Chatbots for Providing Help

Some chatbots can go even further and attempt to help the customer by offering information from a knowledge base. These chatbots often use natural language processing (NLP) to understand what the customer is asking for and search existing self-service articles to surface them for the customer.

If your customers ask many repetitive questions that can be answered by a help desk article, this kind of chatbot will have an immediate impact on the quality of your customer service. Not only will customers get the answer they are looking for, they’ll get them instantly and at any time of the day or night. Plus, every customer that is helped by the friendly chatbot is one less customer that needs a response from your customer service team. This frees up your team to focus on edge cases and difficult troubleshooting questions - those conversations that can’t be addressed by a robot.

Using Chatbots for Simple Transactions

If many of your customer service inquiries are transactional (ie. What’s my balance? When will my order be delivered?), chatbots can be deployed to handle these.

For example, PVR Cinemas offers an online booking platform for movie tickets. They use a dynamic rule-based bot to ask customers appropriate questions to gather information and find the right tickets for them. The questions remain the same based on the flow set by the company, but the data points change depending on the day, location and what movies are available. Customers can easily book their own tickets and PVR Cinemas doesn’t need to staff the live chat with human agents for something that can easily be accomplished with a bot.

Logistics company Safexpress also use a rule based chatbot for simple transactions like scheduling a pick-up and checking a shipment status. Because they ask customers upfront what they are looking to do, they can direct sales queries directly to a human and resolve straightforward transactions with a bot. Every customer gets exactly what they need with the least effort possible - from both customer and agent.

In the near future, expect to see e-commerce platforms using chatbots to facilitate simple return and exchange transactions where a human isn’t needed. While chatbots certainly aren’t going to replace humans in customer service, they are going to be a big help in simple transactional and informational conversations.

Handoffs between Chatbots and Humans

Every person who has called a bank or a telecom company has had the experience of being stuck in an IVR system pressing buttons trying to find a human, or shouting the same phrase over and over again hoping the robot will understand you this time.

Don’t let this happen to your customers who are interacting with a chatbot. Having a smooth transition between chatbots and humans can help eliminate the frustration customers feel when they think there are no humans to talk to.

First of all, have an easy way for customers to talk directly to a human when necessary. If the chatbot gets stuck and isn’t understanding what they want, connect them to a human agent that can provide more specific, niche help.

When you do transfer conversations to a human, ensure that you keep the context from the chatbot. Don’t ask the customer to verify their account again, or to repeat any information. This is one of the most frustrating experiences for customers to go through.

Finally, track what questions are confusing your chat bot - many programs will automatically include this as part of their reporting and insights dashboard. Is it because you don’t have the right knowledge base article created? If so, be sure to update the information that the bot is pulling from. If it’s because the customer had a difficult question that you wouldn’t expect the bot to know - that’s great. That’s exactly where you need a human to step into the loop and help your customer.

Use Cases for AI in Customer Service

Besides chatbots, AI can be used for two other major types of customer service software: reporting and agent assistants. Both use cases keep AI internal and remove the dangers of having robots and AI talk directly to your customers.

Reporting tools

Customer conversations generate a lot of data, and most of that is in an unstructured text format. While most customer facing teams want to use this data, it can be hard to move beyond anecdotal, biased reporting. And that is simply because human brains cannot process that much data. But computers (specifically machine learning AIs) can. Using machine learning to process your customer data can turn the qualitative conversational information into convincing quantitative data. AIs can look at sentiment over different ticket types, or pull out common trends among all the different conversations. And they can do it much faster than data scientists (although data scientists still come in handy when training the AI and analyzing the output!)

Agent Assistants

Instead of letting your AI talk directly to customers without a filter, use AI behind the scenes to route questions to the right queues and suggest canned responses to personalize, edit and then send to the customer. By making your agents more efficient at surfacing answers and helping the right people at the right time, you can improve customer service - and your customers won’t even know that there are robots involved

The AI Customer Journey

The customer journey is a representation of all the touchpoints your customers have with your brand. An AI customer journey shows all of the potential touchpoints where AI can improve your customers experience. For example, say that you offer live chat on your website - you could potentially deploy a chatbot when customers visit your site. Or, you could be more specific with the touchpoint that you target. If, in your business, chatbots are particularly valuable for customer service, but terrible at converting visitors to customers, it makes sense to only deploy the chatbot on the customer service contact page. On pricing pages or product pages, connect potential customers directly to the sales team to help close the deal.


Designing an AI Customer Journey Map

To design your AI customer journey map, first look at all the touchpoints your customers currently have with your brand. Then, identify the touchpoints that could be improved by automating some aspect of the interaction - whether it’s through immediate answers from a chatbot, or triaging questions faster.

You’ll find these touchpoints by looking at two metrics: customers that are unhappy with the speed of service, and places where you see a high volume of the same questions. When you see these two factors in the same place, you’ve identified a prime candidate for AI automation. Another common place to involve AI is in triaging feedback. If you have a high volume of customer feedback about your product or service (either through customer conversations or on review sites) and want to uncover some meaningful data from them, AI can help do that.

Know Your Customers

Once you’ve identified points where AI could help improve the customer experience, it’s time to take stock of your customers. Are they receptive to self service? The odds are pretty good that they are open to finding an answer without talking to a human. 91% of customers say that they would use a knowledge base if it answered their questions. 73% of millenials actually expect a company to give them the resources to solve a problem on their own.

This is where AI-enabled chatbots come in handy. They can improve the effectiveness of your existing knowledge base by making it easier for customers to access what they need. Instead of just searching for what customers are asking for, they search for what customers actually mean.

If your customers are not receptive to searching for their own answers, or aren’t the type to initiate a live chat session, consider if it’s possible to use AI in other ways to improve the same touchpoint. What can you automate to reduce the effort customers spend to resolve their issue? You can keep the personal, human touch in every interaction, but make it more scalable for your customer service team.

Measure and Evolve

Even though AI learns over time, it still requires some human oversight to make sure it learns in the right way. Humans need to be the gatekeeper to a great customer experience.

There’s many ways we can do this - but the easiest is by asking customers what they think and tracking their actions after they interact with a chatbot. This helps open up the “black box” of AI - the idea that we don’t always know exactly how the AI is operating or how they understand us. By measuring the customer experience that customers receive, we can start peeking inside the black box and making tweaks to the process to ensure that every customer’s AI journey is appropriate for their needs.

Customer Satisfaction

After every interaction, either with a human or a chat bot, it’s common to survey customers to see if they were satisfied with the experience. This is called a CSAT survey and is usually a scale of either two options (good, bad) or five (1-5 stars). By comparing how the customer’s rate their interactions with the chatbots to how they rate their experience with human agents, you can see if automating answers is impacting the happiness of your customers. Comments can also be helpful in deciding if it was the chatbot that impacted the rating, or a different issue altogether.

Customer Actions

Sometimes what a customer does is more important than what they say. So even if your customers say they want to talk to a human, they might actually not mind when helped by a chatbot. The only way to see whether your business is actually impacted by deploying chatbots is to measure the behaviours that impact your financial metrics. This means measuring customer loyalty through conversions, churn rates and product usage.

Determine the actions that you want to track based on your desired customer behaviours. For ecommerce, that might be tracking the number of conversions or check-outs per visit. For product support, you might want to track the number of customers that churn after using a chatbot, or where product usage changes.

Evolving chatbot deployment

Once you understand how your chatbot is impacting the user experience, you can tweak the settings to improve it.

For example, if customers with billing questions are consistently unhappy with their experience being served by a chatbot, try removing the chatbot flow from the pricing page. Or, if a customer says they’ve got a billing question, connect them immediately to a human agent.

If the AI is suggesting articles that aren’t relevant, you can remove them from the AI’s view. Using Freshdesk’s chatbot, you can choose which folders the chatbot reads and sends to customers.

Focus on Quality over Cost

When deploying AI, it’s extremely important to approach it from the perspective of improving the quality of the customer experience, and not decreasing the cost of customer service.

It’s true that AI can save your organization money through reducing the incoming volume of customer conversations that need a human to handle them. But companies will see a bigger return on investment from the technology if they don’t only decrease the bottom line, but also increase customer loyalty and revenue.

If you deploy chatbot as a band-aid over not hiring enough staff or not training teams well enough, customers will not stick around. Instead, focus on where chatbots elevate your brand.

How to launch AI and Chatbots in your Organization

Once you’ve decided where to deploy AI and chatbots, how do you get from idea to action? There’s four more things to put in place before pressing the go button on your new smart chat assistant.

Ensure Knowledge Base is Up to Date

Chatbots usually work by serving up existing articles from your help center. If you don’t have well written, easy to understand, current help articles, the chatbot will only be surfacing these to your customers. The first step of investment in AI must be to develop a thorough knowledge base.

Understand your Metrics

In order to track the success of your chatbot and AI strategy, you need to have a way to measure the impact. If you aren’t currently tracking the metrics that you need, get them set up and running before you deploy a new initiative. Otherwise, you’re running blind.

In particular, tracking customer satisfaction, the number of incoming queries you have, and your customers revenue based actions (purchases and cancellations, etc) will help build a basis for proving your hypothesis.

Build your Processes

Chatbots and AI can only work within the constraints of the processes you already have set up. If your existing customer service process is chaotic, adding an additional system will only complicate matters further and you’ll end up with a mess. Always think process before automation. Confirm that the process works before speeding it up. In particular:

Develop your chatbot’s tone and voice

Your chatbot should have a personality. The tone and voice of your chatbot can either make customers feel confident and taken care of... or abandoned to an unfeeling robot.

First of all, make it clear that the customer is dealing with a robot. When a human is involved, make it clear that the customer is talking with a human. Separating the two helps set the right expectations. Blurring the line between robots and humans leads to customers always thinking they are dealing with a robot and not fully engaging in the conversation.

Consider your brand voice. Are you playful and fun? Confident and direct? Professional and efficient? Matching your chatbots voice to something your brand would actually say helps customers feel at ease that they are still dealing with the same company they trust.

Walk through your chatbot’s script as a customer would. Are the right expectations being set? Do they know what format they need to respond in? Do they get all the necessary information from the flow?

Take the time to get the script, tone, and voice right - because many of your customers will experience is!