Conversational AI Best Practice Guide

Here is a short, sharp Conversational AI best practice guide. Conversational AI best practice

First in our Conversational AI Best Practice Guide – Consequences To Your Decisions

When designing a Voice Bot or a Conversation AI interfaces you must know that throwing something together on the fly is not going to work. You must think carefully about what you are going to automate. If you get it wrong you will impact Customer Experience and and may drive many customer to get frustrated.

Start With Data

Before you start to build, look at all the data available and make sure the journeys you are automating are justified based on the numbers.

Scope Out With The Customer

Sit down with the project sponsor to understand what the goals are of a Conversational AI voice bot. Also sit down with employees and understand what their pain points are, watch how they use systems and document the types of conversations.

What Does Success Look Like?

Before a project starts, document what success looks like and how it can be measured.

Examples can be as follows:

  1. Did the customer end the interaction on an error?
  2. How does the caller move around the Conversational AI interface?
  3. Did the caller phone back in?
  4. Did the caller complete the task (containment)

Graceful Error Handling

When designing a conversational AI interface, ensure you trap all types of errors correctly. For Instance:

  • What happens if they say something unexpected?
  • What happens if the customer does say anything?

The goal here is to design in such a way that the customer does not even know an error has happened!

Additionally, always make sure your error handling is on brand. Try and make the error appear as seamless as possible and maybe, where possible provide examples of what the platform is expecting to hear.

Consider your full audience

Does the verbiage make sense to all ages tat are calling into the Amazon Connect Conversational AI voice bot? do you need to consider dialect or how words are used in different cultures.

Playback your Journey using Amazon Polly

Once you have mapped out your journeys, plug them into Amazon Polly and listen to what they sound like.

Use Contractions

Contractions are a shortened form of a word. In Conversational AI this will sound better and more fluid to the caller.

Conversational AI Best Practice – Role Play

After you have tested your journeys out in Amazon Polly, meet with colleagues and role play the conversations to test they sound natural and give the desired approach.

Flow Charts Don’t Work

Flow charts are great for mapping small or specific journeys. Perhaps the main conversations however to map out all the journey in a Conversational AI platform will become extremely messy and therefore ambiguous.

Use Story Boarding

Story boarding is a great way to document the journey in a conversational AI platform. This is a much better approach rather than drawing out flow charts.

Conversational AI Best Practice –┬áBrand

When designing your Conversational AI bot, think about the company brand.

Conversational AI Best Practice – Persona

Persona is important. When scripting the word around a conversational AI platform consider using the training material you would perhaps give to a new start and how you would want a real Agent to respond to questions. Also look at how the company engages over social media.

Finally, in our Conversational AI best practice guide – Empathy

Your Amazon Connect Conversational AI Bot must be empathetic. When handling responses sometime “That’s Great” is used as a response is recognised with high confidence, but make sure these responses are not used when the sentiment of the customers input is negative. For Instance:

System: Welcome to CompanyA, in a few words how can we help you today?

Customer: I’m really unhappy with the service I am receiving and I wish to cancel

System: That’s Great!



If you liked this blog please check out the following:

Beginners Guide to Conversational AI