Skip to content

By Ankur Banerjee on

So, what is a chatbot anyway?

If you’ve used Apple’s Siri, Google Assistant or Amazon’s Alexa—or at least come across them—you may have already heard of the next supposed 'big thing' in artificial intelligence: chatbots.

There’s lots of hype around them, and they’re being touted as the future of marketing and even your next best friend.

Artificial intelligence of this kind is often also accompanied with a reference to HAL 9000, the villain in 2001: A Space Odyssey or Netflix’s Black Mirror, and with stock photography (usually from a Will Smith film).

So with my obligatory references to artificial intelligence in movies out of the way… Chatbots—what actually are they?

To give the simplest example, a chatbot is much like the text/web chat you see offered as online help. Instead of being backed by real humans, they are powered by artificial intelligence engines flexible enough to understand ‘natural language’.

That’s a fancy way of saying that you should be able to describe in your own words what you require, rather than having to press buttons on a keypad until you reach the department you need/throw your phone against the wall in a fit of rage.

Chatbot diagram

Why use them?

Building a system like a phone menu becomes difficult and unwieldy when you’re trying to envisage every possible outcome.

In the past few years, there has been an explosion of the general virtual assistant (of the likes of Siri and Alexa, which want to handle any question you throw at them) and specialist chatbots (the kind that might be handling customer care for a specific company).

Imagine a scenario where you’ve ordered an item for delivery to your home. There’s a ‘sorry we missed you’ card left behind, and you need to rearrange delivery. Most of the time you don’t need to talk to a human, as long as you can find a quick and easy solution to your task.

A lot of the time though, people end up calling a phone number or talking to a person on text chat for these reasons:

  • You could use their mobile app or website to rearrange delivery, but finding the right place to do this can be difficult, especially if you’ve never tried to do it in the past, or if their website or app has been redesigned. We’ve all been there, and it’s really frustrating.
  • If you need any clarifications, diving through an FAQ or help page isn’t very quick and sometimes doesn’t give you the answer you want.

In both scenarios, it’s hard to figure out beforehand if the task will be quick and easy, and therefore worth your time. This is why so many people fall back to the default position of asking someone else to figure it out for you (in this case, a human at the other end of an email/phone call/text chat).

This is where the promise of applying artificial intelligence in this context kicks in with the ‘what if’ factor.

The ‘what if’ factor

What if you didn’t have to wait, and could explain what you wanted to achieve in the same way that you’d explain it to a person?

This concept has given rise to some pretty cool ideas in the chatbot space, with some of my favourite examples below:

Cleo

Cleo is a chatbot that lives on Facebook Messenger as its home (but not run by Facebook itself) that connects to your credit cards, current accounts etc. in a safe and secure manner. Cleo has bank level security and gives you an overall idea of what you’re spending.

Finance tracking apps, including ones that connect to multiple bank/card accounts have been around for a while, but have always left me thinking: I have a pie chart of my spending. Now what?

Cleo flips the idea of finance tracking apps in two significant ways for me:

  1. It may not be pretty, but it’s fast and I don’t need a separate app for it.
  2. By moving beyond just pie charts, it can proactively offer insights/trends that it can then text to me.

DoNotPay

DoNotPay was set up to allow a quick way to contest parking fines, if someone felt they were invalid. This is a great example of chatbot use, since challenging fines usually has to be done via a lawyer, or by writing a rigidly structured letter which most people aren’t familiar with writing.

Crucially, the fact that the output is standardised but the input doesn’t have to be makes it perfect for using AI tech.

In just under two years, DoNotPay has helped users overturn 160,000 incorrect parking fines. The creator of this chatbot has since then gone on to develop a chatbot that offers simple legal aid to refugees, and a chatbot that automatically monitors for flight/hotel price drops and gets money back if they do.

Fully-fledged legal advice, asylum support, travel agencies and so on are obviously much more complex matters, and are beyond what DoNotPay can handle for the foreseeable future. However, this underscores precisely why a chatbot can be useful when it has a fairly specific focus.

In my job at Accenture Liquid Studio, I work with chatbot technology on a daily basis and I’m responsible for the security architecture across projects that use multi-channel chatbot frameworks. I also won first prize at the Accenture AI+VR Hackathon for developing a VR chatbot assistant for helping users manage stress and anxiety. On a personal level, I’m interested in how we can use this technology to improve access to healthcare, particularly in remote areas and developing countries.

Chatbots. The best friend you never realised you had.


To find out more about robots and AI, be sure to visit our Robots exhibition before it closes on Sunday 15 April.

Leave a comment

Your email address will not be published. Required fields are marked *