I had an early morning date with Ana today. While the date was going well, all of a sudden I remembered — Medium? Quora? SHIT!!!
And then I had to cut it short. Ana doesn’t mind. She is quite understanding that way.
Who is Ana?
Ana is the chatbot I have been working on for the couple of weeks, and I am really hopeful of completing her — to the stage where she is ready for her first public appearance — by the end of this weekend.
So, what happened?
Well. I have been talking about how chatbots seem to be too bot like. And I have expressed my displeasure and discomfort with that. If you can not — to some degree — replicate an intelligent, coherent, and most importantly, an intuitive conversational flow, your chatbot won’t work. Users may come to interact with it once, if you are lucky, but they won’t have any motivation to come around for a second time. And just like in any business — repeat and engagement are the two metrics that matter a lot if you are eyeing revenue/sales.
So, I have been working with the flow of Ana again and again. I trashed a few early prototypes, made innumerable changes in a couple, and gave myself a splitting headache every single day.
It came to a point that the best working version of Ana looked complex AF!
Now, I never for once expected the flow to be simple. I am working with a tool for chatbot dev, don’t have to write a single line of code, and it all works as a breeze. If I want to have functionalities into it, it would get long and there would be a lot of building blocks inside. So I was fine with it. But the flow should be simple, isn’t it? At least in theory?
And that’s what the past 48 hours have been all about. I have been thinking on how to make the flow better, slightly less complex, and more meaningful. Last night I think some headway was made in the right direction. And then I slept off.
On waking up, I picked things up where I had left them off last night. And that’s when I got carried away — completely forgetting that I have a daily routine to stick to, and deadline to meet.
Lucky for me, I still have 45 mins left and 1 task is almost complete.
Anyway. Let’s get back to chatbots.
Just look at some conversations I had with chatbots last night.
- 10 million+ likes on facebook
- One of the largest tech companies to come up in recent past
- And Uber couldn’t get one thing right — Language support.
What do I say to that? From the looks of it (and whatever little I remember of my trip to Krabi and Phi Phi), it looks like Thai. But I can’t say with certainty.
Okay. So it was Thai, and it was a bot set up on Uber’s Thailand page as well. Screenshot attached.
But even so, I don’t think that completely puts them in the clear.
- I reached the bot’s page via a featured section on the botmaker’s website (but that’s not Uber’s fault)
- Even if it is the Thai page, one simple check on the user’s locale could have avoided this scenario. And yes, that support does exist.
- If I am from Thailand, continue with the flow. For every single user from anywhere in the world — Show one static message. In English.
A bot that’s truly stuck. God!!! I hate that line by now.
JPARK ISLAND AND RESORT WATERPARK
No contextual message flow? I get it that your hands are tied when using the no-programming bot making tools, but this isn’t so much about the tool as it is about you. You need to think of the possible (or at least the 5 most probable) responses/questions that you would hear from a user at any point of time. And you need to have a better answer to that than your default message.
Take this example itself. I could see something about restaurants in the menu carousel that just passed me by. So I asked about food. Is it unrealistic to expect a better response to that?
Fine. I got a default templated response. So I looked at the particular menu item. Found the phrase “dining options” used there. So I decided to include that frame in my question. And it worked. Then I decided to ask about breakfast — primarily because while I could see the words lunch and dinner used a couple of times in the dining options menu, I could not see breakfast.
And that one question broke the bot again.
THE BACK MENU
I understand that we have been used to the usage and overusage of icons — especially since mobile apps became a thing, but you have to at least ensure what you are doing looks “at least okay!!” — which in this particular case it doesn’t.
AMY WEBB BOT
This was the only decent chatbot I talked to last night.
- The flow was guided
- There was a personal and soft approach to it
- When stymied, the bot didn’t froze, but ‘handled’ it
We really need to start focussing on making chatbots better. Sure, we are moving away from the point where the users WERE FORCED to follow the flow we wanted them to follow, to a more conversational architecture. But unless we start thinking of making bots that at least try to behave human, chatbots would be reduced to ‘just a fad’, instead of being one of the most remarkable supplement tools we have seen for businesses in recent times.