9 hard truths chatbot developers need to learn

Bots are here. Platforms are growing. Developers are shipping products. Conferences are being held. Speeches on the future of bots are being given. VC money is getting poured into the sector. Some exits have already taken place. Brands are getting on board. Picks and shovels are being made. The installation phase for this new paradigm is accelerating with great speed.

At the same time, a lot of interesting areas have yet to be explored and are still underdeveloped. If bots are indeed the new paradigm, the newness is not just confined to their coolness or the messaging context. While it is early in the evolution of bots, it is important to think about concepts that will be crucial for building an actual business in the long run. The buzz around bots is generating plenty of excitement, but it’s the boring and serious topics that will ultimately determine the fate of this space. Lets dig into some these topics:

1. The major players won’t sit idle

Bots built on top of an underlying platform will always be prone to some level of existential risk. What if Slack and Facebook just take the concept of your bot and convert it into a native feature? Ultimately, these big companies own the platform and the distribution on it. Bots that do not threaten the native features of the underlying platform or that don’t operate in areas that are within the underlying platform’s scope and mission are most unlikely to face an existential risk. But regardless of how careful we are, there is always some risk.

2. Nothing is ever free

App developers are used to the “Apple tax.” Will we now have to adjust to a “Slack tax” or a “Facebook fee”? There is no such thing as a free meal. Distribution tax will be a part of the cost of doing business, along with a certain loss of control. While it’s in the underlying platforms’ interests to be developer-friendly, it’s important to remember that these platforms are there to make money.

3. Decide if you have a bot or a feature

Does the bot need to be big enough to warrant a standalone product? Is it small enough to not become a threat to the underlying platform? Having a bot as an additive element in the overall experience provided by a company via websites, apps, physical stores, offices, etc, is one story. But being a bot-first or bot-only company is another. The latter is a tempting and yet risky venture. From what I can recall, app-first companies rose to power years after apps became a thing, following the introduction of the App store. For the first couple of years, businesses thought of apps as an extension of their websites and as something purely additive. We have tons of bots that are merely an extension of other software and brand experiences, and I can’t help but wonder whether going bot-first this early is a good idea.

4. Monetization is key

While it’s easy to think a “bot-as-a-service” business model could be a good way of making money, I wonder whether the business models of the app and website paradigm will translate well to the bot paradigm.

5. Bot bundles are coming

The first generation of apps in the U.S. focused on unbundling bloated web software and websites. They focused on doing one thing really well. There was an app for everything. But once we realized that there are only so many apps we can use, we witnessed a gradual rebundling of apps. Facebook is the best example of a big bundler. Will we go through the same cycle of unbundling and eventual rebundling of bots?

6. Voice bots, text bots, visual bots

When we say “bots” we could be talking about textual bots or voice bots. These types of bots are inculcating different type of behaviors, and they don’t seem to be complementary in the long run. Which type of bots will go mainstream first? (If we open up the definition of bots, we could include visual bots here, as well.)

7. New users finding bots

The problem of finding new bots seems to be more nuanced when the bots are invisible, as with voice bots like Alexa. How exactly do we discover them? Textual bots face discovery issues, too — increasingly so with each platform announcing the many thousands of chatbots they harbor. Besides discovery of the bot’s existence, how do we discover what commands are compatible with different bots? How do we fully know what and how to ask of these bots?

8. Developing the standards

We still need to figure out all sorts of standards and best practices around ethics, security, permissions, privacy, and design. This will take its natural course as the industry moves forward.

9. Fragmenting the OS

While we have two dominant mobile operating systems, we have several bot operating systems. Each has its own perks, limitations, and distribution. It almost feels like picking an operation system in the bot world will become as difficult as making hiring decisions. For an enterprise product, choices are mostly limited to Slack and email, at the moment. For consumer products, there’s a whole host of platforms. For a bot to achieve global domination, it looks like it will need to have a presence on most platforms. The problem of fragmentation raises interesting technical problems of code interoperability, high system maintenance, and need for niche technical expertise. Amir Shevat has taken a great step in the direction of interoperability with his protobot.

I do not claim to have concrete answers to any of these questions. Oftentimes it’s not about finding answers but about exploring questions. I would be happy to chat and get your thoughts on any of these topics. Tweet me at @sarthakgh

Happy botting!

A lot of my ideas have sprung from my conversations with Scott Belsky, Chris Messina, and Amir Shevat.

This post originally appeared on Medium.