Insight Topics

Is your chatbot broken?

By ian tomlin | Jun 30, 2021

Is your chatbot broken? Cover Image

Living with bots

In an era of digital and an always mobile, always online audience, the ‘idea’ of chatbots as a vehicle to supporting customer and prospect inquiries appeals to business owners. Done well, it offers another way of helping audiences to serve themselves with the answers they are looking for. It means any business can economically offer a 24/7 provision to answer support requests, raise interest in new products, and generate new business opportunities.

For many organizations today, the ‘idea’ of bots is where it stops. Getting bots to work and making them useful often isn’t straightforward. People ask a lot of questions, and they don’t always ask them in the same way. Add to this the complexities of language, dialects, anecdotes and metaphors, slang and humor–and suddenly, you’re asking a lot of your bot to get right.

If your business already has a bot, you’re not alone. But remember, every time your bot fails to give the right answers to enquirers, they assume it’s broken. Rarely do humans give technology the benefit of the doubt. Unfortunately, early release of a badly working chatbot producing poor experiences can reduce the inclination to use it.

Back in the day, curious humans spent most of their time online with bots trying to break them. Technology enthusiasts and industry professionals would ask simple challenger questions like “What color are your shoes?” knowing full well bots don’t have feet. All this to prove what they already know–they’re talking with a bot that’s coded by some unfortunate IT guy in the back office who’s being harangued by his boss to make the bot more useful.

Fortunately, the digital audience is getting more used to the possibilities of bots and what to expect. Research feedback suggests that people expect less and less for bots to be ‘human’, setting the bar lower. They don’t need the bot to be human, they just need to know when they’re talking to a chatbot, and know how to answer their questions in the ‘right way.’

Industry 4.0 concept. Big IT specialist robot with hand wrench and small robotic cyborg. Welcome to the new economic future message on blue display. Orange wall background

In a study of thousands of individuals, CEB found that transparency is ranked as the highest priority for consumers, and has the largest impact on loyalty. Trying to pass off a chatbot as a human agent—no matter how well-intentioned—can damage the customer relationships you’ve worked so hard to build.

Right now, most enquiring customers and prospects would settle for a bot that is useful. But what is the issue with bots? Why are they so broken?

Is it the quality of the bot itself, or the training it has received?

Most bots work in a similar way. They provide an empty vessel that programmers can use to hard-code questions and answers into. This is a time-consuming and tedious process. Thankfully, with AI technology, bots can learn on the job. The more they are used, the more often they satisfy inquiries and the more they learn. But it’s a slow burn, and in the meantime, enquirers suffer from bots that are useless at their role.

The way I think of it, newborn bots are akin to Saturday sales staff in my local HiFi store. It’s easy to blame the teenager standing aimlessly at the door for not knowing their products and their sales role. It comes to something when you know more about the product than the salesperson. You’ve taken more of an interest in the topic and done your research before stepping into the store. Then again, we all started out in our careers at some time in the past, and we know how little we knew about the world, equipped as we were with a narrow lens on how commerce and customer service work, and how difficult –sometimes rude, other times intentionally awkward– customers can be.

Why bots give the wrong (or no) answer

When it comes to bots being useful or useless, there are two fundamental types of ‘broken.’

Missed Intents

By far the most common challenge bots have is grasping what humans are asking for in the first place. The potential of questions is limitless. Naturally, hard-coding ALL of those questions and answers isn’t an option, which is why smart bots need to think for themselves. In the industry of bots and Natural Language Processing (NLP), the term used to describe occasions when bots don’t get what you’re asking to do is called a missed intent. The usual response in this situation is for the bot to apologise for not being able to help and recommending a string of useless articles and a link to a customer service agent; to commence a transfer to someone with a brain that rarely works smoothly. At this stage, the enquirer truly knows that your bot doesn’t work and they’re less likely to try using it again.

False Positive

Another form of ‘broken’ is a false positive. This occurs when a bot thinks it knows what you’re asking, and doesn’t– resulting in the wrong answer being given. In some respects, a false positive is worse than a missed intent because the returned data clearly exposes the fact that the bot doesn’t know what the requester is talking about.

Creative design robot on white background. Red head robot toy with empty blue screen, copy space

Fast track your chatbot training

In our opinion, a poor bot is worse than no bot. The longer the training period drags on, the more difficult it’s going to take to rebuild enquirer trust. What you need is a chatbot that’s instantly worldly; a natural linguistic genius with a passion for serving customers. Hard coding answers and publishing new responses to unmatched intents isn’t going to shorten this timeline.

The solution is a learning engine that rapidly tools up your bots to be useful. That’s where a company like INTNT.AI comes in. Our patent-pending AI-driven NLP methods help bots understand missed intents and false positives, to correct them in a matter of weeks, not months or years. INTNT-ENGINE is trained to spot common interpretive errors and recommend resolutions to IT teams that can be instantly adopted with one mouse click. It iterates new learning so that chatbots constantly improve, cutting call transfers, and exposing more opportunities and sales leads.

See INTNT.ENGINE in action. Book an online demo with us or request a video demo