Wizu has several different ways of using Natural Language Processing (NLP ) in order to validate, probe and prompt your customer for more information.

All of the options in Wizu are no code and some are even no config!

Why use AI / NLP?

It's important to ask what we want to achieve: do you want to have a full conversation about anything? We started off doing that and it was clear respondents wanted to talk about everything except the topic we were asking about!

We think of NLP as reverting back to the days when feedback was gathered in person or over the phone with an interviewer. The purpose of the interviewer was to

  • validate the response to make sure it was legible and usable

  • listen to responses

  • probe for more information if required

  • prompt for other topics

It's also important to point out that validated comments hugely improve the Text Analytics accuracy (Garbage In, Garbage Out).

There are 4 ways of using NLP in Wizu

  • Create an AI enabled survey in Wizu

  • Create a SmartComment question

  • Use question Skip Logic

  • Feedback Map

We'll go through each option in more detail...

Create an AI Enabled survey

This is first option to choose as much of the work is done for you and is applicable if you are using a primary measure such as NPS, CES or CSAT. If you are not using a Primary Measure, this option is not for you and you should skip to the SmartComments option.

Create a new survey and choose 'Automated AI'.

This will create a survey with a section with the Primary Measure and a series of follow up questions.

While you can modify these questions, we suggest that this is left to advanced users only.

That's it, prompting out-of-the-box without any code or config!

Here's what it looks like to the respondent

SmartComments

SmartComments is a flexible real-time validation and prompting system that asks questions as the respondent types in their comment.

You can either choose the 'SmartComments' Comment type when creating a primary measure survey, or drag and drop a SmartComment question in the question designer.

The respondent will see something like this:

The prompts underneath the comment box 'nudge' the respondent into providing more details. The level of detail required can be set by choosing the Smart Comments level of effort.

SmartComments can be used in combination with the Automated AI option when choosing a new survey in the Survey Builder.

Use question Skip Logic

Another option is use custom Skip Logic to prompt for more information.

Available conditions are (common conditions are highlighted in bold)

  • Contains word stem

  • Contains exact word or phrase

  • AI: Entire answer sentiment

  • AI: Single entity sentiment

  • AI: Stage sentiment

  • AI: Touchpoint sentiment

  • AI: Mentions this touchpoint

  • AI: Mentions this Stage

  • AI: Parent Topic mentioned text

  • AI: Parent Stage mentioned text

  • AI: Topic not mentioned count

  • AI: Stage not mentioned count

  • AI: Owner aspect count

  • AI: Owner touchpoint count

  • AI: Check follow-up prompt type

  • AI: Check number of entity mentions in conversations

A really useful condition is to check the sentiment of the comment and use Skip Logic 'AI: Entire answer sentiment' to check if the value is > 0.3 (reasonably positive) or < 0.3 (reasonably negative). If they are very negative 😠 use a different follow-up question than if they were very positive 😍.

Equally you can listen out for particular word derivatives using 'Contains word stem' so for example you have some logic to end the survey if the respondent mentions any derivative of 'danger' such as 'dangerous', 'dangerously' or the route word 'danger'.

Feedback Map

Feedback maps give greatest control of what to probe for and what not to probe for.

They will be covered in a separate article, and there is some configuration involved with creating a Customer Journey Feedback map. If you are starting with Wizu, we suggest starting with the other options first and then comeback to using Feedback Maps if you need to.

Did this answer your question?