Why your restaurant needs a bot that actually works

If you’ve been listening to any of the tech giants [1][2][3] recently, you would’ve noticed a trend. Bots. Bots everywhere. All of a sudden, all of your favourite messaging applications support bots — from Facebook’s Messenger, Skype, Telegram, and Slack. It’s exciting that so many platforms are supporting bots, this creates some serious opportunities for businesses to proactively reach their customers in their personal lives, and work lives. However, bots aren’t new — the world has already gone through a wave of bots. For example, if you pick up the telephone and call your favourite err tolerable telecom company, you’ll probably be greeted with a half-human half-psychotic-robot that will prompt you to select a menu item from the tens of items that are available. Now, I haven’t done any extensive research — but I’m certain no consumer likes those automated systems.

This is why we need to make bots different. They need to be precise enough to make things convenience for your customers, and they need to be conversational and expressive enough to delight your customers. If we can create quality bots that can achieve both of those things, we can truly replace apps.

In order to achieve those things, we need a natural language service that is focused on one particular task. It makes perfect sense that in order to create smart bots, we should create bots that are focused on a certain area. For example, if you run a restaurant, you want your waiter to be really good at serving your customers — answering questions related to your menu, taking orders diligently, making sure that the experience for your customers is top-notch from when they enter your door till they leave.

The other major feature that bots provide is a platform for your customers to be expressive. Living with a vegetarian roommate, when we decide what to eat, we usually perform the same rituals. Open Google. Search restaurant. Look at Yelp. Is there a menu? There isn’t. It’s somewhere on the restaurant’s website (which by the way, barely works on my mobile phone). Now to find what vegetarian options they have. Wait — what’s in this moong dal dosa? If I’m allergic to something, I’m probably out of luck — I need to pick up a phone, and talk to real person to find out if I can eat something.

Now imagine this — talking to my friend on Telegram:

Me: “@DosaBot do you have any vegetarian options?”

DosaBot: “We have the following options: Moong Dal Dosa, Paneer Masala Dosa, Vegetarian Thali, …”

Roommate: “@DosaBot nice, never tried the Moong Dal Dosa before, whats in it?”

DosaBot: “It contains split mung beans, basmati rice, coriander, salt, green chilis, and corn oil”

Roommate: “@DosaBot, awesome! So no onions in there right?”

DosaBot: “The Moong Dal Dosa does not have onions”

Me: “@DosaBot, what time do you close today? Can I make a reservation for 2 tonight?”

DosaBot: “We close at 10pm. Hold on while I make that reservation for you for 8pm tonight”

A conversation like this would take a few seconds and would provide more information then the search on Google would.

Sound exciting, doesn’t it? But here’s the catch — NLP is hard [4][5][6]. That’s why we need to focus creating bots that are trained for specific tasks — like a normal worker would be. This is the idea behind Kedo — a service that can help you build your restaurant bot and publish them to multiple channels all at once.

The downfall of apps isn’t that they aren’t intuitive enough, its just that there are too many that just don’t work and live up to the expectations of the consumers. Let’s not make bots that way too. Let’s make bots that are truly smart, convenient, and allow your customers to be expressive.

[1] — http://time.com/4291214/facebook-messenger-bots/

[2] — http://www.torontosun.com/2016/04/13/skype-bots-change-branding

[3] — http://techcrunch.com/2016/04/12/telegram-beefs-up-its-bot-platform/

[4] — http://rainbird.ai/nlp-is-hard-2/

[5] — https://zompist.wordpress.com/2016/01/08/why-nlp-is-so-hard/

[6] — https://www.quora.com/What-makes-natural-language-processing-difficult