What happens when you build a bot in 5 minutes?

February 13, 2018

Here’s what I learned building a drag and drop bot

There are a lot of companies out there promising that you can build and launch a bot in “{insert number here} minutes.” And many of them guarantee that you can do this without coding.

Doesn’t this sound too good to be true?

That’s because it is.

As someone whose programming skills are less than worthy to work for a tech company, I’ll be the first to admit that I would love to build a bot in a short amount of time and without coding. But, I’ve come to realize that this doesn’t work for all use cases. The build and deployment may only take a few minutes, but you’re going to need to sink a lot of time into improving the user experience, and may hit the limitations of the system.

Don’t believe me? Keep reading.

Before continuing, a quick note on use cases that rapid bot building platforms are very well suited for:

  1. Publishers: Allow your viewers to subscribe to your content and search through content to find posts that interest them.
  2. Marketing: Build a subscription list and send them promotional content such as discounts.
  3. Rapid prototyping: Create and test bots to know if your customers will respond well to bots and to understand the capabilities of bots.

In the rest of this post, you’ll see what happened when I set a 5-minute timer and started building. I explore what can be accomplished in 5 minutes as well as the limitations of these types of tools.

What can you do with a “5-minute bot”?

Let’s start off with taking a look at what I ended up with after building for 5 minutes.

I created a simple bot with a leading “5-minute bot” provider. You can try it out here.

Getting into the bot, you’ll see that the bot I’ve built can carry on a simple back and forth conversation and can send the user any amount of information that I want. The use case that I’ve put together would be great for a content publisher who has built up a Facebook page following and wants to update their followers when new content comes out.

A user can subscribe to notifications when new content comes out, or just take a look at the available material when they interact with the bot.

I’m also able to accept input from the user as natural language, making it seem like the user is participating in an actual conversation.  

Finally, I can use a menu to answer frequently asked questions.

Overall, I'm pretty happy with how quickly I built a content publishing proof of concept. The RSS feed was much easier to implement than I anticipated, and adding natural language was more than I thought I would be able to do in 5 minutes.

We’ve seen the capabilities of a “5-minute bot”. Let’s dig into what we’re missing in this bot.

User experience

There are a few places in my bot with a limited user experience. I’ll give you five minutes to try to find them. Here’s the link again.

Go ahead; I’ll wait.

Found something that you didn’t like? Here are a few things that I found frustrating to solve or potentially impossible to fix:

  1. Identifying missed natural language utterances - If you said something that my bot didn’t understand, you’d see a whole lot of nothing. Or, it may have misinterpreted what you said, leading to a confusing conversation. As the developer, I can’t identify what you said that the bot failed to understand or fix a misinterpretation. With more time, I could have trained for more utterances, but the tool limits my ability to figure out what statements I should add.
  1. No personality - This is primarily because adding a personality to a bot takes a lot longer than 5 minutes. Although I could have done this, limiting myself to explore what could be done in 5 minutes made this impossible. PS: In this blog, I discuss why a bot personality is critical.

One of the highest priority goals for anyone building a bot should be to enable an incredible user experience. Achieving this goal was definitely impossible in 5 minutes, but I’m also hitting some limitations of this “5-minute bot” tool.

Natural language processing

Many of these “5-minute bot” platforms have built-in natural language processing (NLP). NLP training is arguably one of the more tedious steps to building a bot and not having to do this would be a dream come true.

Hold up. Why don’t you want this?

Complexity. Your company has in-depth domain knowledge that your customers will refer to within a chat. Having NLP training done in the background prevents you from finding and addressing mistakes when they come up. Without insight into the NLP training, a customer may ask a simple question, with wording slightly different from what you expect them to ask, resulting in the bot falling flat on its metaphorical face.

Debugging

“5-minute bot” platforms have been designed to be intuitive and easy to get going. However, they have left out an essential part of development. Something will inevitably not work perfectly the first time, and some debugging will be required.

Debugging is not easy. I do see red dots when I forget to add a link, or something is wrong with my block. But when I’m testing on Messenger and don’t receive a response from the bot, there’s no indication anywhere of what happened.

Complex business objectives

The term “bot” has a pretty negative connotation. The reason for this is simple. A lot of bots out there are not intelligent. At best, the average bot mimics a personality and can handle a few simple commands.

Although these bots may have mass adoption and complete their designated tasks, they aren’t doing anything to move artificial intelligence forward. So, how do we accomplish this ambitious mission?

Developers are critical. Limiting ourselves to simple conversation logic will prevent exploration of the limits of artificial intelligence. Let’s focus on empowering developers rather than removing them from the equation entirely.

Developers enable your bot to tie into your business back end and logic. Developer focused tools are also platform agnostic, allowing you to build one bot and deploy it to multiple channels. Finally, developer tools enable an extensible bot with an API and webhooks.

With the increasing popularity of chatbots, it’s no surprise that companies are looking for ways to make them quickly and without a substantial financial investment. But building a bot using a “5-minute bot” tool makes it much more unlikely to provide the user experience that your customers deserve. These tools serve a unique purpose in allowing you to see if your audience will interact with chatbots, but you may find that you need to upgrade to something more to truly engage and support your customers.

How about another option? At Meya, we believe that developers are a critical part of every cognitive application. Enable amazing user experiences while accelerating your roadmap with Meya. Learn more and sign up for a free trial here.

Rachel Pautler

Helping developers build bots as smart as they are at Meya.ai

You might also like