Bots Are Failing And Here’s Why

By Andy Peart On Jan 2, 2018

A new chatbot pops up almost every day. Facebook, for instance, claimed 100,000 unique bots for messenger back in April 2017. But despite their number, many of them are failing to impress, leaving users underwhelmed and frustrated.

But why are so many of them failing to impress? There are five key things to keep in mind when answering this thorny question.

You need a brilliant bot or your users will try it once and never come back. It has to be an end to end brilliant solution & it needs to be conversational!

One of the main reasons bots are failing is their conversational ability – or to be more precise, lack of it!

Brilliant bots must give users a humanlike experience, therefore if you’re building a bot, you need to make sure it is able to truly handle natural language conversations. Many of the bots out there may work fine if you follow a linear, prescribed way of engaging with them, by asking the questions and providing information in exactly the way that the bot developer intended. But unfortunately, we as humans tend not to do this—we communicate using natural language. We branch off at tangents, we circle back, we miss out crucial facts and figures, we ask for clarifications. We want to be able to speak to bots in a humanlike manner, and we want those bots to understand us.

In short, we want to have a conversation.

Now you’re getting lots of data – how can you utilize it to improve your business processes?

When people communicate in a natural, conversational way, they reveal more than just the words they’re saying; their individual preferences, views, opinions, feelings, inclinations and more are all part of the conversation. When many thousands or even millions of conversations are captured and analyzed, this data becomes a unique and powerful source of customer insight whose value can be further enhanced when it is cross-referenced with other sources of knowledge. The challenge is to understand what this data means and what effective actions to take as a result.

Recognizing new or unexpected trends while staying in tune with customer behavior and sentiment is a competitive advantage, and a key to building successful customer relationships, maintaining loyalty and increasing repeat business.

Make your bot memorable; give it a persona that reflects your company’s brand values

It’s a question that comes up time and time again: do you need to give your bots a persona? My short answer is if you’ve spent millions on building a corporate brand and positioning for your organization why would you ever consider not doing the same for your conversational applications?

The longer explanation is a little more involved.

For example, two airlines, British Airways, and Virgin Airlines. Both fly across the Atlantic, but each brand has a completely individual and unique persona. Not just in their uniforms and the branding on their planes, but in the way they portray themselves with every interaction.

I’d argue that personas within conversational interfaces are as important as your domain content, the knowledge your bot uses to function. It’s the personality that elicits the most reaction—good or bad, from the customer. Get it right and you’ve created the opportunity to improve the customer experience, even more so than just giving the correct response.

It doesn’t need to be overtly funny or extravert, just reflect the brand. Taking the time to consider some of the finer points in how your natural language application will respond and react verbally early on in the project will pay dividends later.

The underlying model must be a hybrid approach so that customers have a seamless experience speaking to the bot online from a desktop or mobile device.

It’s a common misunderstanding that machine learning systems somehow work completely on their own, without human supervision. Nothing could be further from the truth. Just as linguistic-based conversational systems require humans to craft the rules and responses, machine learning requires humans to collect, select, and clean the training data.

I would argue that the ideal approach is to combine the best of linguistic and machine learning models in one solution. This flexible approach allows enterprises to quickly build AI applications whatever their starting point – with or without data – and then use real-life inputs to optimize the application from day one.

You’re not building for just one channel, it must be a conversational system that is deployed across many channels, so an omnichannel approach to conversational intelligence!

As an enterprise, which platform and channel are you going to bet on: will it be Alexa or Google Home? Will Messenger be the channel of choice? What role will the web, mobile, even wearables play? The truth is, your customers want to use them all – at different times and for different reasons; but they want a consistent experience that ideally remembers what has been said in previous conversations.

The solutions are for enterprises to be able to build conversational intelligence once, and then deploy across multiple channels, platforms, and languages.

Ultimately, if chatbots don’t deliver a true conversational interaction, one that delivers contextual understanding that is consistent across different channels, in any language, it will be a failure for your enterprise. Chatbots require features that can provide a true conversational interaction because it allows for a great customer experience. It also helps to increase brand engagement, because you never know what a human will ask next!