Insight of the Week: What’s the best way to get started with Conversational AI?

by?Kerry Robinson

I ran a LinkedIn survey over the last week asking: "What's the best way to get started with conversational AI?"?

The results were interesting. And a little concerning...?

The majority, 51%, said you should start with Strategy & Vision first. The next most popular option, by far, was to build and learn as you go. Just a handful would advocate designing & testing with real users.

I was surprised that Design & Test performed so badly. It got me thinking: why such reticence to design and test??

I can think of three possible explanations:?

1) Lack of prototyping and test tools?

There are lots of tools for prototyping web and mobile experiences. But far fewer, and much less mature tools for chatbots and IVR. This is a real shame. My team have been designing and testing conversational experiences for more than two decades. And we still learn something new. Every. Single. Time. Granted, there are some experiences that we know so well that the ROI of prototyping and testing is weak, so we go straight from design to production. But these are the exception. Not the rule.?

You need to be resourceful. Find a way to prototype and test with real users. If nothing else, sit back to back with someone else and role-play the design. But that has its own issues. There are better ways to do it.?

2) Over confidence in design?

Just because we can talk, doesn't mean we know how to design. And just because a designer can design, it doesn't mean they know how real users will react to their designs in real-life situations. Only prototyping and testing will give that insight.?

There are lots of new designers entering this space. Hoovering up best practice guides and getting experience on a few projects. But how well are those designs working? Really? Platforms don't make it easy to understand the experiences customers have with our designs, so it's easy to design, build, deploy, and think everything is OK.?

Prototyping and testing designs isn't a sign of weakness. It shows strength and maturity in your designers and design process.?

3) Lack of time and/or resources?

Prototyping and testing isn't free. And it takes time. But there are ways to do it that emphasize speed and rely on the 80:20 rule. No user testing means 100% of usability issues make it into production. The cost of finding (or not finding) these issues in live should not be underestimated: customer experience and brand impacts, incorrect routing, under-utilized self-service. Reduced containment.?

Find a way to mock-up the key paths through your IVR or chatbot design and test with real users. The feedback of 5 or 6 users that interact with a realistic prototype can weed out 70-80% of these issues.?

That's a strong ROI right there.?

I'm delighted that my LinkedIn audience have recognized the benefit of Strategy and Vision. And of building and learning as you go (as long as?you've got the right data!). But I think we're missing a trick if we don't design and test experiences with real users.?

If we’re not already connected on LinkedIn, let’s do it, so you can see what me and others in this area have to say… and hopefully become part of the conversation. Reach out and connect with me?here.?

Kerry

Kerry Robinson?is an Oxford physicist with a Masters in Artificial Intelligence. Kerry is a technologist, scientist, and lover of data with over 20 years experience in conversational AI. He combines business, customer experience, and technical expertise to deliver IVR, voice, and chatbot strategy, and keep Waterfield Tech buzzing.

Subscribe to Kerry's Weekly AI Insights