his year, the tech industry has lost its collective mind over bots, and travel companies quickly joined in the mania. As of July 1, there were already 11,000 bots on Facebook Messenger, and thousands upon thousands of articles on the so-called bot revolution. Admittedly, I’ve been slow to embrace the trend, especially after a few tests ended in an endless loop of frustrating, non-sensical conversations with a robot. So, when a friend bet me that a bot couldn’t be created in a day, I jumped at the opportunity to build one. Partly because I am always up for a good challenge. Partly because I wanted to create a better alternative for getting industry news than my overly cluttered RSS feed. And mostly to try to better understand the hype around bots.

The result: mark.ady bot (pronounced mark-ah-dee), a content-driven bot on Facebook Messenger that serves up marketing insights as well as curated travel and tech news from key trades, based on a user’s interests. It also provides users with a few fun ways to kill time when they need a break from their busy workdays (don’t worry, the bot won’t tell your boss). mark.ady bot taught me some valuable lessons about bots. Here are a few of them:

A Basic Bot Is Simple To Create (Seriously!)

One of the reasons that bots have gotten so much hype is because they are said to be easy to create, particularly when compared to developing a mobile app. As someone with limited programming skills, I was skeptical. But, with the help of free tools readily available to help beginners like myself, it was actually really simple to get a bot up and running.

If you are thinking about building a bot, I highly recommend using a “BaaS” (bot-as-a-service) tool. mark.ady bot was build with the help of Chatfuel, a free service that allows you to create bots on Messenger or Telegram with little to no coding knowledge, but there are other tools, like MobileMonkey, that also offer great platforms on which to build.

Remember Why Bots Are So Hyped: Value, Ease of Use, Speed

While creating a bot is simple, developing one that is actually worth talking to requires quite a bit of foresight and planning. First, if people are going to use the bot, it needs to offer a clear value (the value that mark.ady bot provides, for example, is customized, curated industry news). The bot must then provide that value in a way that is faster and easier than if a user were to download an app or visit a website to perform the same task. With mark.ady bot, a user can tell the bot his or her interests and receive news about that topic from all key industry trades immediately, which is much faster than visiting the websites of each trade, opening dozens of daily email newsletters, or culling through an RSS feed reader.

Interactions Are The Bot’s UX

With bots, there are no fancy interfaces. No shiny objects. And nothing to hide behind if you create a lame bot. The entire user experience will live and die with how well the bot can interact. After having worked at a SaaS company, this felt scary, uncomfortable, and downright weird.

Because interactions are so important to the success of a bots, it is important to spend some time planning out the flow of the user interface and trying to anticipate all possible interactions that users can have with the bot. This way, your bot can react accordingly, and it keeps users from getting stuck in a confusing, messy loop like this:

Photo from TechCrunch: https://techcrunch.com/2016/04/13/facebooks-new-chatbots-still-need-work/

When building mark.ady bot, the majority of time was spent on this planning, and I would have spent much more time on this important part of the development process, if I hadn’t been under the one-day time constraint.

There are plenty of other things to consider when thinking about a bot’s interactions with its users. Here are a few that I found useful:

• Onboarding: Even if you think your bot is intuitive, the bot should offer a quick intro in its first message(s), including info about what the bot can do and a suggested first task for the user.
• Personalization: Try to personalize bots as much as possible to make the experience more intimate (ok, I didn’t do much of this due to time constraints, but you should).
• Visuals: Use graphics or video to enrich the experience, wherever it makes sense.
• Testing: Test early, test often, and test with anyone willing to help. It is really helpful (and surprising!) to see how people interact with the bot as you build it (more on that later).

Buttons Are Golden

When building a bot, it is easy to get carried away with AI to interpret all interactions. Resist the temptation. Many times, buttons are the best way to interact with users. This is because they are fast, easy to use, and more accurate because they don’t rely on your bot to interpret user inputs correctly.

When buttons aren’t a possibility (e.g. if your bot facilitates a hotel booking), then be very clear about what you want when asking users for information so that the bot can correctly identify it and react.

Then what about the fun, artificial intelligence piece? mark.ady bot’s AI is layered in to interpret user’s interests and to make sense of variations of commands (e.g. users can get to mark.ady bot’s menu by typing menu at any time. mark.ady bot also understands similar phrases like: main menu, features, main, see menu, home, etc. to get users back to the menu). The bot is also working on some small talk, although it is still learning the nuances of natural language.

Bots Need A Personality Too

Do you like talking to that guy in your office with the personality of a wet cardboard box? Didn’t think so. Don’t let your bot be that guy. While mark.ady bot is indeed a bot, it still has a personality. A bot worth talking to will keep the conversation light, personal, and stick with the tone of its brand. If you are working on a bot and happen to be a horrible writer (hey, no one’s perfect), then recruit your marketing department or hire a copywriter to draft up some snappy messages.

Studying User Behavior Is Crucial

Optimization is key for any product, particularly bots. While the optimization of mark.ady bot began only after my day-long quest to launch a bot, the learnings have been significant and are worth mentioning. It is an understatement to say that users did not interact with the bot as I had anticipated. So, the bot has undergone many facelifts since it was created, including new AI, new functionalities, and, most significantly, an entirely new menu. All of these changes were implemented based on analysis of user interactions. The menu, for example, was originally laid out horizontally so that users had to scroll right to see each menu item. It turns out, this was far too complex, and, as a result, users  missed some of the bot’s key features. After simplifying the menu, users flowed through the bot’s paths completely differently, with the data suggesting that the bot became much easier to use.