They say you learn more through failure than success. After building bots for a few years, I'd say I definitely agree.

While I've had a head start, I know many businesses are thinking about bots for the first time.

That’s why I want to share ten things I wish I knew before I started. Learn from my mistakes and build better Facebook Messenger bots, sooner.

1. Do Fewer Things Better

Spend more time planning than building. Figure out the single most impactful function you can expose or create through Messenger.

When I first started building bots, I got carried away with all the exciting things that I could do. I didn't spend enough time focusing on the one thing that I should do. Remember, bots are a tool. You need an business application and strategy to put them to use.

2. Conversational Copywriting is Different

Talking to our friends is natural because we do it every day. It's much harder to zoom out and break down the art of how we message our friends.

We use more than words to express ourselves. Messenger supports a variety of expressive mediums, like emojis, stickers, and gifs. In 2017, users shared over 1.7 billion emojis every day on the platform. My friends say about 5% of them came from me. Figure out what's the most appropriate and delightful way to engage with your users.

3. Realize Who’s on Messenger (and Who Isn’t)

In my early days, I didn’t realize the sheer number of people that user Messenger on a regular basis. With over 1.3 billion monthly active users, it doesn't feel that far off.

It's a global behemoth, but every country has their preferred messaging app -- and it's not always Facebook Messenger. For example, WhatsApp (technically another Facebook-owned platform) dominates usage in Latin American countries. Before you commit to a platform, take the time to understand your audience and where they spend their time. You want to meet them where they already are -- not force engagement on an app they don't even use.

4. Your Bot Doesn't Need to Be a Genius Right Now

Most companies and journalists in this space have let you down. They’ve set an expectation that all-knowing AI is here -- and that your bot needs to be powered by state-of-the-art artificial intelligence in order to be valuable.

That's bullsh*t. When I first started building bots, I didn't realize that plain old IF/THEN logic trees are actually a great way to build a Messenger bot your users will enjoy engaging with. If you can read a "pick-your-path" adventure book, you can build a bot.

5. Don’t Deceive Users

Most people don't care who -- or what -- solves their problem. There's a misguided stigma that bots will always be a bad experience for users. That's not true.

It's the bots that pretend to be human that frustrate us. It only takes consumers a few questions to figure out what's going on. Make sure your users know they're talking to a bot from the very start of their interaction, and be clear about what users can and can't expect from your Messenger bot.

6. Solve for Utility, Then Delight

Your bot can -- and should -- have personality without trying to be a person. Conversational copywriting lets you weave in enjoyable experiences at every turn. But delight shouldn't be your bot's primary purpose.

I've built plenty of bots that were fun but provided no value. Those experiences taught me to focus on utility early on, not delight. Solve a real problem first and iterate on the rest over time.

7. You Can't Just Replace Your Humans With Bots

Automating everything would be a bad experience. And handling everything yourself would not be scalable. That's why humans and bots need to work together.

I learned that it starts with thinking about what you're asking the user to do. Bots are great for repeatable and predictable conversations. But when it's a delicate situation, like sales or customer care, opt for people if you can. Bots empower humans to focus more on important conversations.

8. Plan for Growth

In "Field of Dreams", Kevin Costner said "if you build it, they will come."

Great advice for building a movie baseball field, but horrible advice for marketing. The biggest problem with Messenger bots is people finding them in the first place. But it's easy to get more messages to your bot. Start with small wins, like a CTA on your website or Facebook page.

9. Everyone Expects Certain Things

No matter what purpose your bot serves, users always want the ability to unsubscribe, restart the conversation, or talk to a human. There are certain implied things that consumers expect all Messenger bots to be able to handle.

The best way to plan for these is to crack open a thesaurus and make a list of all the ways users might request these core functions. Then make sure your bot can handle them. There's nothing more frustrating for a user than not being able to unsubscribe.

10. It Will Break. A Lot.

Bots are still very new, and there's a good chance that many of your users have never talked to one before. New users might not fully understand the limitations of a bot right away, and they will it's likely that their conversation will start to stray away from how you want it to go.

Error messages are an overlooked part to building great Messenger bots. Spend time on them. Make sure they're a seamless part of the experience.