<img height="1" width="1" style="display:none" src="https://www.facebook.com/tr?id=979905748791482&amp;ev=PageView&amp;noscript=1">
Chatbot ToolsJanuary 11, 2017Written by Dean Withey

Make Your Own Chatbot in 5 Easy Steps [GUIDE]

make your own cha

Five steps to creating your very own chatbot

What will your robot do?

So you're jumping on the chatbot bandwagon huh? It seems as though everyone is making a personal chatbot at the moment. From Chris Messina to Dharmesh Shah.

If you want to get involved, here is a guide to making your personal chatbot.

Step 1: Decide on your messaging service.

Most people tend to build a chatbot using Facebook Messenger. It has a nice open API and clear documentation (as well as over 1 billion monthly users).

But, depending on your location and audience, you might want to look at other options. If you're in, or near, Asia then WeChat, Line and Viber might be a better fit, alternatively, if you're a younger person in the USA you might want to look at kik. If you want a truly global coverage, then good'old SMS is your friend (albeit at a cost).

Ideally, you should deliver a chatbot experience across all messaging apps ubiquitously. Only then will you be able to offer an experience on the platform your user prefers, instead of forcing them to join the platform your service uses (and therefore lowering adoption).

One of the problems with the current ecosystem is the lack of unifying API to chat across all messaging apps at once, in fact, at the time of writing, there are only really two options. ubisend or smooch.

Step 2: What do you want your chatbot to do?

Is your chatbot a way for people to chat with you in real-time? Perhaps you want to broadcast content to subscribers or drip-feed over time? Maybe you're delivering additional functionality to compliment your business?

Regardless of your requirements, you should have an explicit conversion, or goal, for your chatbot.

As an example, my chatbot is designed to collect subscribers and keep them up-to-date with my blog posts and general activities. I use my chatbot a little bit like an email list. I gather people, keep them engaged and entertained, and get an 83% open rate and a 53% click rate... compare that to your email list...

Dharmesh Shah's GrowthBot is different. It's designed to be a tool, in this case for inbound/SEO. His chatbot metrics are probably around adding value to his ecosystem of HubSpot tools. Perhaps with goals like increased engagement, positive feedback and increased brand awareness.

Step 3:To AI or not AI? That is the question.

Generally speaking, there are two main types of chatbots.

Some guide you through an experience using buttons and asks for specific responses. Or, others emulate human interactions and use AI/ML to understand the intent, language and requirements of a user.

Personally, I don't think AI is quite good enough for a seamless conversational experience (yet). We currently, as of Feb 2017, recommend our clients don't try to pretend their bot is human. In my opinion, AI needs another couple of years to be able to deliver a human-level conversation that is not a barrier to the interaction.

Step 4:To be human or honest?

Will your chatbot pretend to be human, or will it flat out say "Hey, I'm just a machine and here to help.".

Pretending to be human (as mentioned in Step 3) is tough. You better be sure you're one of the world leading experts in AI to get away with it (if you are, and reading this, I'm super-honoured!). If your chatbot pretends to be a human and slips up, you've lost your user. People aren't stupid; they'll recognise if your bot gives fluffy answers or keeps saying "I don't understand".

Perhaps you could try to be human, but come clean at the start. "Hi, I'm an automated chatbot, but I'm super smart, and you should just be able to talk to me like a human. How can I help?". At least this way your user will be a little more tolerable when/if the chatbot can't keep up.

Top tip from all the chatbots we have built

We have built many chatbots and this very simple tip has improved the usability and retention of every single one of them.

Early in the conversation, tell your users what to expect. If you're delivering a service and not monitoring manually, say. "Hi, this is an automated service. To speak to a human call this number or reply with the word MESSAGE". This way people understand what is going to happen and know how to get help should they need it.

Step 5: Start building.

You have your channels, goals, needs and type of chatbot organised. Now you need to start building it. We'll talk deeper about this in another post, but for now here are some pointers:

ubisend - www.ubisend.com - All messaging apps and SMS, code-free, pricing based on subscribers.

chatfuel - chatfuel.com - FB Messenger and Telegram, code-free, enterprise level over 100,000 messages.

Botsify - www.botsify.com - FB Messenger, pricing based on subscribers.

Smooch - www.smooch.io - All messaging apps, API-only, pricing based on messages sent.

(Beep Boop - www.beepboophq.com - FB Messenger and Slack, API-only, pricing based on server rental. Beep Boop seems to have gone away, updated on 04/05/18).

DialogFlow - www.dialogflow.com - FB Messenger, API-only, pricing based on messages sent. (Formerly known as API.ai, updated on 13/10/17).

Botkit - www.botkit.ai - FB Messenger, Slack and Twilio - API-only, free. (PS: Formerly known as howdy.ai, updated on 13/10/17).

(Init.ai - www.kitbot.me - FB Messenger and SMS, API-only, based on subscribers. - init.ai was shut down. Updated on 13/10/17).

So there we have it.

A 5-step guide on how to start building your personal chatbot. I hope it was useful and got you thinking about the whys and goals behind your chatbot.

I'll post another article about actually creating your chatbot soon. In the meantime, if you have any questions about ubisend or would like a look at the platform, get in touch!