How To Article

How To Get Started With Chat Automation

How To Get Started With Chat Automation

Rules-based automation and artificial intelligence have expanded the horizons for business processes and continue to alter them. However, taking a step towards chat automation requires different considerations and is overall more difficult to achieve because of the real-time nature of chat messages.

Although the introduction of message automation in your customer chats can provide much-needed relief to customer service representatives, it may also reduce your overall customer satisfaction if implemented poorly.

If implemented correctly, it will allow your customer support staff to take more time to solve issues with higher complexity that require empathy and/or technical experience.

Sadly, the process of message automation is not one that you can just flick on and off like a light switch. It is a process of trial, error, and optimization. Although the road is long, the pay off can be high. Some companies will be able to reduce their customer support costs by orders of magnitude.

Without further ado, here are five steps that will enable you to embark upon message automation successfully.

1. Evaluate Your Customer Conversations

The first step to message automation is understanding the chats your customer service representatives are having with your customers. In this first step, you'll want to understand the potential pay off you can achieve by automating some conversations.

You'll want to find the answers to three critical questions:

  • What are the most common problems your customer support team responds to?
  • How often is the answer to each of those questions similar or the same?
  • How much time does your team spend answering those questions?

In your initial research, you'll want to ask the business development team, sales department, and customer representatives for anecdotal answers to these questions. You'll be surprised how annoying repetitive tasks can easily be recalled by the employees that perform them.

Ideally, you'll want to ask employees who are already participating in chat conversations. As the flow of communication in chat channels can be vastly different from those on email or over the phone.

If at this stage you find your staff saying they don't experience any repetitive conversations, you may not want to embark on message automation just yet.

2. Identify Patterns

Although the idea of chat automation often conjures the thought of employing and all seeing all knowing chatbot to respond to your customers, we haven't yet seen a successful case of message automation that works like this.

The first steps to practical chat automation are relatively simliar to business process workflows. These little improvement might use AI or simple rules to automate responses to messages that come in over and over and over again. Thus, leaving time for complex conversations where humans actually add value.

How are you going to decide which conversations to automation first? Once you've received feedback from your teams, you should have a clue of what to look for and where to dig. Here's a method we use:

  1. Seperate the feedback from your team into topics
  2. Find 50 to 100 chat conversations per topic
  3. Group those conversations by response

Now all you have to do is ask yourself, how many of those questions could you reply to with canned answers? As long as you can come up with at least one canned answer, then you now where to start.

3. Experiment With Automation In A Closed Environment

Before you go ahead and unleash chat automation upon your web visitors, users or customers, you'll want to test out some solutions in a closed setting.

At Rocketbots we usually do live testing in a channel, which doesn't have that many subscribers. For us this channel is Telegram, we have about 100x fewer subscribers on Telegram than we do on Facebook Messenger.

For your first experiments in message automation, we recommend you start in one of two places. This first is an onboarding flow.

For some of our customers, we see that one question or topic is so prevalent that it's best to preempt it for all new subscribers. This means creating an onboarding flow or a survey that can educate your new subscribers about the topic or issue.

The second involves introducing canned replies. These canned answers would be the ones that the automated system will be expected to churn out during a live consumer chat if a subscriber gives the frequently asked question.

Once you've prepared the canned answers, ask your staff to use them in your test environment. With this test, you'll want to look out for two items.

First, how do people respond to the canned response? Does it make sense to them? Do they ask a follow-up question? You'll want to use this time to optimize and test your canned response so that there are as few follow up questions as possible.

Second, you'll need to define a clear distinction between what falls within the scope of the chat automation design and what requires a customer representative to take over the chat proceedings. Even for frequently asked questions, you won't always have a canned response that fits in every case.

In case the case falls beyond the function of the canned response, there must be a clear process of how the chat will be redirected to a human correspondent.

4. Introduce Chat Automation

You’ve evaluated communication patterns, tested pre-defined answers and now it’s time to introduce chat automation. Where should you start?

When choosing where to actually go live, you'll want to consider the following matrix.

This is a chat automation matrix.
Chat Automation Matrix

Make sure to start your message automation where you have a high probability of being right and a low cost of being wrong.

While chat automation can assist in catering to multiple routine queries at a single time, it cannot replicate the human empathy and problem-solving skills that are derived from experience.

For the remaining two types of interactions, organizations can implement a mixture of semi-automation and human correspondence where automated replies are being sent out until either the consumer is satisfied or the chat agent takes over for further assistance.

5. Analyze, Optimize & Implement

Nobody expects you to get it right the first time. Chat automation is a trial and error process that requires organizations to test better variants of their answers repeatedly and analyze consumer feedback to automated chats to improve their design.

The key to a successful chat automation system is the ability of the organization to record, evaluate, and improve, based on their findings.

Iaroslav Kudritskiy

Iaroslav Kudritskiy

Iaroslav is the Chief Operating Officer at Rocketbots overseeing growth and operations. He is a graduate of the MBA program at the Chinese University of Hong Kong and his previous roles include positions at Kodak Alaris, Xaxis & Light Reaction.

Leave a Reply