July 29, 2022 – Duration 23:28

S3 Ep. 2 How not to bot

Bots can deliver great automated service for a wide range of questions, issues and transactional interactions. But they can’t make up for bad customer experience (CX). That might be why customer satisfaction with bots has declined as bot use has grown. Organizations too often jump into drafting requirements and deploying a bot without laying the groundwork for success. But great bot experiences start with strategic planning based on a deep understanding of how customers engage with your brand. In this episode, Jane Hendricks, Senior Product Marketing Manager at Genesys, explains how to create bots your customers will love. And she advises listeners how not to bot. With the right insights, you can build better bots that deliver empathetic experiences — and help customers advance on their journeys.

Read the episode highlights below

Subscribe to our free newsletter and get updates in your inbox on new episodes released each week!

Jane Hendricks

Senior Product Marketing Manager at Genesys

Jane joined Genesys in 2021 to lead Product Marketing for Conversational AI. She brings over 20 years of experience in product marketing, product management and consulting, covering data-driven disruptor technologies like data science, machine learning and artificial intelligence (AI).

TTi20 S3E2 Conversation Highlights

Here are conversation highlights from this episode, slightly edited and condensed. Go to the timestamps in the recording for the full comments.

✖

Give us your insights on the reality of what bots can and can’t do.

Jane Hendricks (02:25):

Expectations for bots are high. Companies will look to bots to do everything from deflecting simple interactions to completely automating and revolutionizing self-service. And pretty much everything in between. And when you have expectations that are so high, there’s no way all those expectations can be met.

It's interesting that you say that. Our “State of Customer Experience” report found that bot usage has increased significantly among consumers, but satisfaction with bots has declined. Considering all this, what are some things to consider when launching a bot?

Jane Hendricks (04:17):

Some bots weren’t necessarily put in place for the right reason or built the right way. What do I mean by that? When you have a positive interaction with a bot, the bot fits into your experience seamlessly. It’s able to understand what you say; it’s able to respond; and it doesn’t leave you with that dead end. The experience is negative when the bot doesn’t help the customer solve their problem or advance on their customer journey.

There’s a lot we can do better to make that experience more holistic, more empathetic. For example, make sure it fits within a journey; that it’s not isolated. Make sure that it’s personalized and is talking with the human in their terms. Make sure there’s continuity; that the bot transitions the interaction to an agent if it can’t handle the request.

What are some of the requirements to consider when launching a bot?

Jane Hendricks (06:49):

People will start with, “I need a bot, and this is what I want my bot to do.” But they don’t think about the bot in the context of the total experience and where would it make the most sense for the customer to put an automatic conversation in place. There are some markers to look for, like multiple entry points. That’s something that a bot can help facilitate.

If you start with determining what’s best to automate and what should stay high touch before you jump into what exactly the bot will do, what operational systems you’ll need, what language models you’ll need to build, etc., that will create a much better experience going forward.

Is consumers’ dissatisfaction with bots primarily due to customers feeling like the bot’s purpose was deflection? Or is it simply a poor customer experience?

Jane Hendricks (09:18):

I think it’s a combination. Some bots are built using keywords, not natural language. They’re not built to be conversational, so it’s not a great experience. A bot that tries to do too much is not such a good experience. And a bot that leaves you at a dead end is a problem… Not starting off with journey mapping and understanding how a bot fits, the odds of success are slim.

You’ve talked about how important it is to test and making sure that you're mapping actions to intents. What should that testing should like?

Jane Hendricks (16:50):

You want to make sure your bot can understand all the ways people are asking for whatever it is you want the bot to help with. And if somebody says something that is completely outside of that, you need to know that it’s being said and figure out what to do with it. That’s the importance of training a bot — and why it’s so critical that the bot fits within the context of the total journey.

How can an organization ensure ease of use or time to value for the business users who are putting the bots out there in the first place?

Jane Hendricks (18:18):

Bots are automatic conversations. The business users and the people who own the customer experience have a sense of what those conversations should look like. So, putting tools to build these conversations into their hands will accelerate the time to market because now they have control over that input. But there are still areas where developers and linguists should be involved; you want the business and IT to work together.