Best Use Cases for Conversational AI, Voice, Video, Etc.

By Kevin Karty

Pros and Cons of Different Open Ended Research Methods

I've been asked a lot about WHEN and HOW to use different types of open ended technology. There are so many options available, but which is best for which use cases?

As early adopters - we did voice & AI before it was the cool thing to do - we've had plenty of experience with all of the modalities. The chart above is my attempt to provide guidance based on this experience. It's not perfect, and in some cases there are nuances, but it's a pretty good first cut.

I have to say that our go-to is AI-guided voice story questions, because they scale so well. I know a lot of the industry does video, and it's great for some things - especially showing product usage or in store shopping. But it's both unnecessary and actually problematic for a lot of other use cases.

Case in point, people don't LIKE going on selfie-mode video for regular surveys. They just opt out. To give you a sense of why, how often do you FaceTime with a stranger instead of just doing a regular voice call? (Or even a text?)

Likewise, there's a lot of talk about AI replacing human qualitative moderators. Some of that will happen, but a lot of the use cases for AI moderated interviews really don't mesh well with AI. All the data I've seen suggests that AI moderated interviews tend to yield much shorter answers than interviews with real humans, and their ability to really explore new or deep or new topics is limited.

Here's my summary of what I think are the ideal use cases (and things to avoid) for each modality:

Ideal Use Cases for Each Modality

Again, these use cases are based on actual experiences.

Practically, we find ourselves using AI Guided Voice Story Questions the most: 

• They scale well

• They have been an amazing tool to prevent fraud.

• The interactive AI voice questions also had a pretty good pull through rate (and a lot of the refusals are likely fraudulent)

• We've used them to great success even in sensitive categories like health related issues. For example, we ran an extremely powerful internal survey on GLP-1 meds for weight loss, with excellent participation and tremendous depth and honesty.

• You can QUANTIFY them easily. We've done studies with literally thousands of 2-3 minute voice response questions embedded in a big quant survey, and coded open ended answers into statistically reliable quantitative data.

I know that AI-moderated video interviews are the "cool thing" - and we do them - but participation rates are much lower (or require higher incentives) and there are a lot of cases where it's really not practical. We tend to like them a lot for things like product reviews or a mystery shopper exercise. The key question is whether the application uses front facing video, or you're asking the respondent to take a selfie video. 

If it's a selfie video, do you really need to see their faces? This is a big deal. Recording faces has a lot of disadvantages - it's a MUCH bigger file to upload on a phone, it has all sorts of additional privacy restrictions (you have two forms of identification - voice and image - so this 100% qualifies as PII and triggers all sorts of security/GDPR issues), and you could lose up to 80% of your respondents unless you pay a big incentive.

I'm not one to sell hype. I hope the info above is useful!

If you have a different view or I missed something, please join the conversation on LinkedIn:

See Post on LinkedIn

Or contact us to set up time:

Contact Us!