Dashbot panel: Top Considerations for Building Enterprise Customer Service Chatbots

Recently, we were pleased to be part of a panel hosted by Dashbot.io about important considerations for enterprise companies looking to build chatbots for customer service. There was a great lineup of panelists from AccuWeather, 1-800-Flowers, Aspect, as well as InTheChat, and the conversation was packed with excellent examples and actionable advice.

So what did the panelists talk about as the top considerations for building enterprise customer service chatbots? Well, first:

The elephant in the room: don’t expect to replace CSRs completely

For all the hype flying around about buzzwords like machine learning, artificial intelligence, and natural language processing, chatbot technology has a long way to go. There are inquiries that chatbots just won’t be able to handle well, if at all. So the best way to look at customer service chatbots is to see them as a complement to human agents rather than a replacement.

The truth is that the strengths of chatbots are different from the strengths of human agents. A chatbot will never get frustrated about getting asked the same question repeatedly, for example. However, bots won’t have the ability to provide the human touch. Nor are they well suited for resolving complex issues that require both analysis and the ability to be empathetic to a customer’s concerns. Ideally, implementing a customer service chatbot should be about balancing the human and the AI. It’s important to reassure your CSRs that you’re not looking to replace them, and that a customer service chatbot will take the load of boring routine inquiries and give them the time and space to use their skills to resolve more complex customer issues.

Ideally, a customer service chatbot should make your CSRs’ lives easier, not harder – which will only benefit your customer experience. Your customers will get better service from happier agents, and a chatbot will give them quick and instant access to basic information, even outside of normal business hours.

Where do you start?

The first question you need to ask is: who is involved in the conversation?

At many companies, the only people in the room when a new bot is being planned are developers. It can be easy to dismiss the process of bot planning and building as a task that should solely be handled by development teams, but building a successful customer service chatbot that will engage with and delight customers requires collaboration from multiple parts of the organization. So who exactly should be at the table?

First and foremost, good chatbot development starts with knowing how people are actually going to interact with it, which means your front-line CSRs need to be involved from the beginning. Having them at the table is key, because they’re the ones who are interacting with your customers! And if you ask a front-line worker what their most commonly fielded questions and issues are, their answers will usually surprise you! So customer service input is critical.

Start with asking customer service stakeholders about their most common inquiries for simple information that could be easily automated. Then follow up by gathering customer interaction data to get an accurate picture of what types of inquiries are being handled, and how often those inquiries occur. Then use that data to define the problems that your bot should address. Without customer service involvement, developers are often left to brainstorm intents themselves. The problem with this approach is that often, effort is wasted developing intents and interactions that customers never actually use, not to mention that some common but non-obvious intents will not be developed at all.

It’s also important to ensure that writing of bot scripts not be left to developers, because a customer service chatbot will necessarily need to be friendly and consistent with the personality of your brand. Chatbot scripts need to be written by people who can write natural-sounding language that conveys personality while still representing your brand professionally. For many organizations, this means having experienced marketers and copywriters handling bot scripts, but some organizations are looking further afield to professionals like screenwriters!

The next question is: What does customer success look like?

There are many reasons and goals for implementing customer service chatbots, so it’s important to ask what does customer success look like for the inquiries the bot will be used to address? Some possible approaches to customer success include:

  • Will customers have return engagements with their chatbot? If they interact with your bot once, will they be comfortable enough with the process to use it again?
  • Do customers who use the bot to address inquiries or service issues need to subsequently follow up on a different channel? If customers are connecting through your chatbot and then need to follow up in a separate interaction through voice or email, then the chatbot has actually been counterproductive!
  • Does the chatbot free up your agents to improve their service on more complex inquiries?

However you define customer success, having a clear definition will allow you to construct metrics that will ensure that your chatbot is a net benefit to customer service and customer experience.

Bot-Building First Steps

Once you have assembled your team and defined what success looks like, the next step is to clearly define the problems to be handled by the bot.

(When looking at problems that can be handled by the bot, make sure that emotions are part of the process. As 1-800-Flowers’ Jennifer Hui pointed out, a sympathy flower arrangement is something you would never want handled by something as impersonal as a chatbot!)

Once you have defined your use cases, your customer service stakeholders and developers can work together to map out the appropriate bot flow for each use case.

The last rule of thumb for teams during bot-building is:

Challenge assumptions and don’t be afraid to stand out

Be aware of trends and consider how the choices that you make about the chatbot avatar will affect how your customers perceive you. For example, as virtual assistants have gotten increasingly common, people have started to talk about the problem of all-female virtual assistants. Playing into that trend can cause your customers to perceive your brand as one that doesn’t value the diversity of its customer base.

An additional interesting example was raised by the panel moderator, who talked about a customer who experimented with different avatars for their chatbot. They found that interactions with the female-avatar chatbot were the most aggressive. Interactions with the male-avatar chatbot were less aggressive, but interactions with a chatbot whose avatar was simply the company icon were mostly neutral – which brings up an interesting point to consider: ultimately, when you personify something like a chatbot, your customers will approach it with all the same baggage that they will when interacting with a real person.

​Lastly: Set expectations for your customers, and make bot handoffs easy

When implementing a new customer service chatbot, don’t try to pretend that the bot is human. Be open that the bot is a bot and be open about its capabilities and what types of issues it can best assist with. Never say that your bot can answer “anything”, because when it fails, your customers will be annoyed with you for making promises that you couldn’t deliver on.

Also, don’t be afraid to make it easy for customers to bypass the chatbot. If a customer wants to speak with a human agent, interacting with your chatbot isn’t going to change their mind, and they’re only going to get frustrated the harder you make it for them to connect with a human. So make sure that connecting with a human is always quick and easy and ensure that the handoff from bot to human is smooth and easy at any point in the interaction.

This means making sure your agents have access to the chatbot flow when the bot hands off to a human agent, so that your agent doesn’t have to frustrate your customer by asking them to repeat information. Contact centers may also want to make taking bot handoffs a specialized agent skill, so that customers who are transferred from the chatbot to a human are receiving the best service.

Of course, this is a pretty high-level summary of an in-depth conversation, so if you’d like to see the complete panel you can watch it here. Or if you’d like to talk to our digital customer service experts about making chatbots part of your customer service landscape, you can click here to contact us.

You May Also Like

About the Author: Anna Kreider