June 1, 2021

Best practices to set up a bot from scratch

If you want to implement an artificial intelligence solution for customer service, or if you already have but want to improve the experience you're offering, here are some tips we always give our customers when they start working with a conversational bot.

1.   Identify who will move the project forward

In our experience, we suggest differentiating between two key roles:

• Project leader

The one with decision-making power. They must be available to participate in weekly project follow-up meetings.

• Knowledge specialist

The one in charge of consolidating the conversational bot content. That's why this person must have great writing skills to create questions and answers that are empathetic, solve customers' needs and are in line with the company's tone and voice. They should be available part-time for this task.Technological knowledge and customer service experience is a nice-to-have, but they aren't essential requirements to implement a bot.

2.   Determine your knowledge base

In this step, you identify frequently asked questions, come up with the best answers, and organize them.

• You can start with 20 questions. Choose the ones you get every day from customers. Don’t worry if you don’t reach that amount. Over time you'll be able to expand your bot's knowledge.

• Organize the questions on a spreadsheet. As a guide, here is the spreadsheet template we use with all our customers.

• Define the tags for your questions. You should include them in the knowledge spreadsheet. They'll help organize your content by topic.

You should also include the answers to these questions on the spreadsheet. Creating the best answers isn't rocket science, but there are definitely some things to consider. The next part of this article specifically addresses this issue.

3.   Design the experience

Now it's time to create the answers that the conversational bot will give to your customers. Here are some key tips:

• Choose the tone and voice of your bot. Figure out how your bot will speak; if it'll be colloquial and casual, a little sassy or sarcastic, or if it'll use more formal and reserved language.

A word of advice? Think about your customer service agents, their age, gender, personality, how they speak, and any other relevant trait to characterize the bot. With this in mind, create the answers as if this person were the one responding.

• Think about how complex the answers need to be. This will depend on the audience interacting with the bot: will the answers be technical or addressed to the average user? Will your customers be the main audience or will it be targeted for internal teams?

• Create short, clear and specific answers. Remember that writing answers for a chat won't be the same as writing an email or web page. Avoid cutting and pasting information without adapting it to a conversational format. This is important because it'll make the customer experience more conversational and natural.

If your answers are long, we suggest dividing them into different shorter topics and recommending related answers using complements to complete the information.

• Don't forget the greeting, goodbye and thank you messages. They're crucial in providing a personalized and natural experience. You also have to think about the answers the bot will give when it doesn't understand questions or when it doesn't have information to answer them (at Aivo we call these "evasive" answers).

Related article: The power of evasive answers for effective communication

It's also really important to consider what the bot will say when it gets insulted. Even though we'd love for it not to happen, the bot may receive the occasional unfriendly message. Make sure to have some witty answers on hand for these situations.

• Upload more than one answer per question. This step is key to "personanality" messages, just like the ones we were discusing above. Having several options for a greeting or a thank you message, for example, will make the bot sound more natural.

• Avoid answers that start with “Yes” or “No”. These answers won't sound right when the query isn't a question.

• Answer customers question on the channel they used to contact you. The idea is that the virtual agent is the first resolution channel and its goal is to solve the problem. If most answers only transfer the client to other channels, this becomes an annoying step before getting to customer service, which impacts on customer satisfaction and customer service cost for the company.

Your conversational bot is waiting for you

In a nutshell, you don't need to be a computer or programming specialist in order to get a conversational bot up and running. With these tips, any team from any company can get great results.

If you still have questions or if you'd like to talk to our team about conversational bots and artificial intelligence, you can schedule a quick call.

Are you looking for new ways to improve your CX?

Our customer service solutions powered by conversational AI can help you deliver an efficient, 24/7 experience  to your customers. Get in touch with one of our specialists to further discuss how they can help your business.