Betaworks, a White Star Capital partner, recently hosted its Botcamp day. Betaworks CEO John Borthwick, my fellow Kellogg alumnus Matt Hartman, and the rest of the team put on an excellent day with great speakers and demos by eight promising bot startups.

Here are my 10 key takeaways from the day’s presentations and discussions:

Current state of affairs

  1. Early days: Bot functionality is still limited. App developers are laying early groundwork to improve user experience, bot personality, data access, algorithm development, NLP (natural language processing), and much more. The bot world today is developing apps that are far more useful than Clippy, yet we know that the industry is still in its infancy and that bots don’t do everything users want them to do. For a fun example: Even though you can order a Domino’s Pizza directly from Facebook Messenger, the bot’s functionality is fairly limited — you certainly can’t ask it if pineapple will taste good on your anchovy pizza.
  2. Naysayers: Many people used to argue that they would never purchase products online. Now many of those people loathe going to stores and prefer the convenience of purchasing from the comfort of their homes. A similar transformation of people’s preferences will likely happen with bots, especially as bots continue to improve, with access to more data, better algorithms, and the evolution of best practices.

Product strategy

  1. Asia: A lot of people point to China’s use of bots in WeChat and expect that the same bot products will work in the Western world. This is not necessarily a fair comparison. Setting aside the greater cultural and societal differences, we can look at a very specific technological/political example: Many Chinese developers rely on existing app platforms as a way to bypass the requirement for government approval for apps (SAPPRFT: State Administration of Press, Publications, Radio, Film, and Television). Such limitations do not exist in the Western world, and so bot developers are not as reliant on messenger platforms as they are in China. Secondly, Chinese users entered the social web via an all-encompassing messaging platform, QQ, then WeChat. Western users have been educated to think about “atomized” apps and generally use a wider mix of messaging platforms.
  2. The two rules of marketing: Many moons ago, one of my business school professors hammered home the point that businesses exist only to either a) address a pain point, or b) delight their customers. The same advice holds true for today’s bot developers, except that at botcamp they often state it as: a) “help people get shit done” (X.ai’s Amy, Olabot’s enriched profiles) or b) “make it fun” (Hugging Face’s virtual best friend, 2001’s Subservient Chicken).

Thoughts on functionality

  1. The New Deeplink: Bots are great for jumping into the deeplink of a decision tree. Compare a bot to a telephone IVR (interactive voice response system). In an IVR, you listen to several options, then hit 5, more options, then 7, more options, then 2. A bot can directly route you to the right department when you say just one word (“cancel,” “support,” “stolen,” “reschedule,” etc).
  2. ♩ ♪ Lean back, lean back ♫ ♬: As bots continue to improve, we should expect to see users spending less time leaning in trying to figure out how to get the computer to do what they want. Instead, the bots will become so good that people will lean back and simply say what they want in their own natural words. We are in the Age of Context. An example of a recent improvement: Ask Siri to tell you a joke about Donald Trump, and she will give you Google search results. Ask Alexa, and she’ll make you laugh.
  3. Apple iMessage: The once-simple messaging app released in 2011 that 250 million people use has recently undergone a massive change. iMessage now supports third-party apps. While the press has discussed this to a certain degree, the impact of iMessage is still widely underestimated. Expect iMessage to continue evolving and to become an important platform for bot developers, especially in the U.S., where iMessage has its largest foothold. And iMessage will transform to include more elements of fun for a younger audience.
  4. Dream bot (one of many): Imagine being able to chat with the author of a self-help book. The chat could be powered by a bot, but users wouldn’t even be able to tell the difference. Imagine feeling like Stephen Covey, Dale Carnegie, Malcolm Gladwell, Jim Collins, and The Dalai Lama are all ready and eager to offer you advice.

Some challenges

  1. Discovery: We do not have a good model for bot discovery. It is hard to find the best, most useful, most fun bots. At the same time, there are many potential privacy issues to resolve, especially in open environments (unlike Slack’s closed environment). For example, if I can invite a bot into a group discussion, the bot’s functionality might erode some people’s privacy expectations. This is one of the reasons Facebook is moving slowly with bots: They represent a big opportunity, so Facebook doesn’t want to make a wrong move and alienate people forever. Even with recent progress —  Facebook is rolling out an ad unit that takes users from their News Feed to a conversation inside Messenger and is adding support for payments inside of bots —  Facebook is being deliberate and careful.
  2. ♩ ♪ It ain’t easy ♫ ♬: Being a bot developer is not easy. In addition to the challenges of discovery, user experience, algorithms, getting access to the right data for AI & NLP models, etc, the platforms also impose their own limitations. For example, you cannot pass parameters to Facebook Messenger that would allow the bot to jump into the right location in its algorithm. Another example: Some data, including the types of KPIs that VCs often ask for, is not always made available.

Conclusion

While the bot industry is still in its early days, bot developers have already created many useful bots. Some favorites discussed at the event: YourMD’s health insights, Xiaoice’s best friend (with its 20 million registered users), Digg’s freshest news (White Star is an investor), Swelly’s surveys, Poncho’s weather, Digit’s money saver, Purple’s political news, Talkspace’s therapist, Toronto-based Zoom.ai’s workday assistant, and many others. These are exciting times for bot developers and for their users. Even with all these entertaining and productive bots, the best is yet to come!

This article appeared originally on Medium.