Rise of the Robots

I have to say, when I first heard Facebook’s announcement about ‘chat bots‘ being the Next Big Thing ™, I was a tad skeptical. However, as the dust begins to settle, I may well be changing my mind…

I was cautious because there is so much, so much hype surrounding all of this technology, that’s it’s easy to get confused & right to be wary. Perhaps it’s best if we take a step back to try to see the wood for the trees…

Messaging Apps

A huge % of us on Planet Earth have used SMS text messaging, and the smart phone evolution of texting are messaging apps; there are a few out there (WhatsApp etc) but Facebook Messenger is what we’re talking about here (see below).

Screen Shot 2016-05-06 at 19.33.47

Many of you will recognised this, (well, not this conversation as it’s between me and my sister-in-law), but you’ll recognise the interface and the concept: a small window where you can chat back and forth to one or more people. So far, pretty easy.

The situation gets funkier when we consider this:

Could we interact with brands/companies/services via chat?

The answer is obviously yes. Which is why there are plenty of services like Zopim where you can allow your (e.g.) website visitors to chat directly to your operators in realtime.

One benefit to the customer is that the chat window is very free format – they haven’t got to learn how to navigate a website etc, they just type in what they want and the operator will (hopefully!) provide an intelligent response.

But does that scale?

And that’s the issue. Do we really want to connect every customer directly to a human operator? Well, maybe yes in the ideal world, but we’re also mindful that automation can save time and money; so we wonder if there is a way for technology can help here?

….and this is where the bots help.

Are these the droids you’re looking for?

Instead of having (expensive) humans be the front line, first point of contact on (say) Messenger; can we have a ‘bot‘ (i.e. an ‘intelligent‘ agent, a computer program) service customer needs instead?

The answer is…..it depends (cop out!).

The issue here is Artificial Intelligence is still pretty crummy; yes we can teach a computer to wipe the floor with the world’s Go champion, but can the same computer handle a loosely structured conversation about which tennis shoes you want? I doubt it.

But are we setting the bar too high?

Yes, we probably are. I don’t think the goal here is to create programmatic experiences that trick us into thinking there is a human being at the other end of the chat box: no, the goal is to create an easy way for customers to make a meaningful, positive interactions with your brand: set expectations and people will understand the constraints of the medium.

Exhibit A

One of our clients helps people find work. One aspect of their business is that workers can register the hours which they are available for work – this obviously makes it possible for the client to match people to current resource requirements.

Currently, registered workers submit their availability via a private form on the website. The website is responsive, the form has been designed to work well on mobile; so, whether the person is on their laptop or their smartphone, they can keep the availability database updated (and keep themselves eligible for work).

But what if we could make it even easier for the users to supply this availability information?


One way  could be an app; a very slick mobile app which not only runs on iPhones but also runs on Android etc (after all, we don’t know what smartphone the audience uses). Such an approach is fine, but let’s be clear here: a non-trivial, smartphone app is not cheap to develop & maintain; and a non-trivial smartphone app which runs on multiple platforms is even more expensive. Ouch.

And what about when you find a bug in your app? Well, you create a new version, and you’ve got to ask all your users to update their version to this one – do you have a headache yet? Quite. And an expensive headache at that. Apps are great – but they come at quite a price.

Piggy Back

But what if we could avoid a lot of that expense?

What if we could piggy-back on top of an app which is either already installed on a huge amount of phones (and available on a plethora of devices). Better still, the development & maintenance of the app is not your concern.

Who is this generous app developer you may ask?

A. Facebook

Or, to be exact, the Facebook Messenger app.

So, if they are happy to provide us with the platform, why not use it?

The Solution

The client could have a chat bot app for their registered workers. As you need a Facebook account to use Messenger, that makes it easy to work out who is a registered worker and who is not.

But what can these ‘registered workers‘ do with our mythical Facebook chat bot app?

Well, they can fire up their iPhones and, via Messenger, they can interact with the chat app with messages like this:

I’m not available to work tomorrow

I’m on holiday from the first two weeks in August”

I can work next Tue, Thr, and Fri in the afternoon

Or they can even interrogate the chat bot:

What is my availability for next week?”

“Where am I due next?

When is my next free day?

…and the chat bot can respond accordingly.

Sounds impossible?

No, no it’s not. Think for a moment. You’ve used interactive voice/phone systems for years. This is the evolution of that.

Is this easier than submitting the form on the website? Again, it depends. Depends what device you’re on. Choice is the key here. If you’re out and about, firing up Messenger to say “I’m not available for work tomorrow” is probably a darn sight easier than manipulating a web form on a small smartphone screen.

Programming the Conversation

Indeed, Facebook themselves offer software to allows you to construct conversations and responses. So, as long as you are clever enough with being able to parse what people are saying, and providing them with options – you are on to a winner.

And remember, this ‘parsing‘ doesn’t have to be omnipotent; if it doesn’t understand what you mean all the time, it can say when it is confused. And it can present you with further options. Like an adventure game of old. Remember this?


Above is screenshot from the old Spectrum adventure game of The Hobbit (yes, I am the at that old). For those of you too young enough to remember, Grand Theft Auto this wasn’t, your input was the text prompt: you typed commands and got responses. E.g.



..but you can




…you get the picture.

Explain to the user what the vocabulary is, and give them options so they can achieve their goals.

Artificial Intelligence

As touched on earlier, Facebook have been kind enough to create Wit.AI to help with the creation of your chat bot conversation:
Screen Shot 2016-05-06 at 21.10.20

What we’re seeing here is a clever, graphical tool for creating the ‘scripts’ which manage this back and forth conversation between the user and the bot. And, if your interaction is even more complex, involving your own database look ups of availability (or products, or songs, or whatever), then you can roll your sleeves up and write your own business logic.

Closing Thoughts

The cost of developing a useful chat bot app is certainly not trivial; but at least you are spared the cost of developing the actual mobile app as well (that’s not to say downloadable apps are never valid, just making a point here).

Hopefully that’s given you some food for though on the the Chat Bot front. What are the opportunities for your business?





Tags: , , , , ,
No Comments

Leave a Reply