How Artificially Intelligent is a Chatbot?
5 minute read
James Fox profile photo
James Fox

2017 is set to be the year that artificial intelligence turns the world on its head. Widespread automation will lead to a loss of jobs in every sector with transportation, logistics and administration as the first to go. Donald Trump will take over the world at the head of a droid army, which is likely to look like the Trade Federation from the dreadful ‘Star Wars: Phantom Menace’. But first things first, throw away your marketing plans and replace everything with chatbots…

At least, this is what we keep hearing. But no one really knows what will happen, or indeed when it will happen. Despite the undeniable technology improvements, stories of pizza delivery robots, and a succession of humiliating defeats for board game masters, the reality is that in many cases the technology is still experimental and bears little resemblance to the one used by brands in their pizza chatbots. Even the fairly common assumption that automation will mean jobs are lost is a matter for debate [soft paywall].

Disentangling chatbots and machine learning

We’re certainly not suggesting we’re cynical about the potential for artificial intelligence now or in the future. Our suggestion would first be to try to draw some clear distinctions between true machine learning and the types of chatbots and other robotics which are currently more prevalent.

Machine learning, especially where neural networks are concerned, is incredibly complex. The skills required to undertake these kinds of projects are rare, as this fascinating story of how Google Translate has been transformed by machine learning highlights. Even with their vast resources, Google are still coming to grips with the field and face strong competition from Facebook and others.

Chatbots on the other hand are generally far simpler, relying more on data and decision trees - which isn’t to say they can’t be useful. Integration with messaging services such as Facebook Messenger and Kik means you can get instant access to millions of users.

Getting the best from chatbots now

At the moment, chatbots are most suited to performing tasks where you have a clear decision process that can be followed. Uber’s Developer Experience Lead, Chris Messina, sums it up quite nicely for us below. Rather than considering chatbots as standalone services, or even products in their own right, think about this:

“What is the service that you’re building and how do you create and support the relationship that you have with your customers through messaging?”

You can read more about Chris’ views here.

In fact, Uber have been quick to determine a clear use case for ordering cabs via Messenger. Their service isn’t intended to replace their hugely popular app – which they’ve spent considerable time on to improve the experience – but to fill an additional purpose. If a user has friends or family visiting, particularly those who aren’t users or who use Uber less frequently, it’s easy for them to quickly book a cab from an airport or station straight via their conversation.

Pizza Hut recently came up with their own way to make use of chatbots. Not content with telephone, web and mobile ordering, they created a new system for customers to order their pizza directly from Facebook Messenger and Twitter. In addition to typing in what you want to order, users can also get details about special deals, opening hours and delivery areas.

Whilst not as PR worthy as Dominos forays into Amazon Alexa Skills (with very limited functionality and negative reviews) or robotic delivery (only a test), Pizza Hut are harnessing more of the potential uses for the technology as they currently exist.

Guided conversation versus human interaction

Passing the Turing test is also probably an area best left to those with experimental neural networks to power their interfaces. But, as Pitchbot shows, a chatbot guided conversation works perfectly to gather answers to a specific set of questions. In this case, an otherwise potentially dull set of data is made into a far more human and entertaining conversation with the addition of a simple bot.

In summary, there are plenty of scenarios in which a chatbot can add a useful layer to your service. Platforms and solutions are also continuing to emerge all the time, reducing the barrier to entry for everyone from large brands to startups. And as with any new and emerging technology, it can be tempting to jump in head first and test the waters, but it’s important to remember that not all chatbots will genuinely improve an experience or fulfil a user need. Our recommendation? Think before you jump - and don’t call it artificial intelligence. You might be setting a customer expectation you won’t be able to meet.