State of The Bot: where are we now and where are we heading?
Originally posted July 2016
A few months ago, I moderated a panel “To Bot or Not to Bot**” at the Samsung Accelerator in San Francisco, bringing together stakeholders from across the bot ecosystem. The three ecosystem POVs represented were the messaging platform (Rod McLeod from Kik), the bot developer (Vijay Sundaram from Polka Dot Labs) and the bot investor (Amit Garg from Samsung’s VC arm). The fact that I moderated instead of a chatbot is a reminder how far the technology still has to go, and my job is safe for at least a few more weeks. Given our audience and panelists, we focused most of our attention on chatbots, rather than some of the integrations/bots that startups have been building on Slack.
Six big themes emerged from the panel (and post-panel happy hour and post-panel happy second and third hours) that have proven even more prescient in the following months as more bot-themed news has rolled out from tech’s big boys (Google, Amazon, Facebook, Microsoft, etc.) as well as a steady stream of investments and updates from the newer kid on the unicorn block, Slack.
1 — The majority of bots and chatbots are currently pretty “meh”. I should learn better than to have the first bot-takeaway be a downer, but it was clear from the POV of the investor, the platform and many in the audience that the current quality of bot-experiences left a lot to be desired. When asked to raise hands, only the rare few in the bot-biased audience had used a bot they found enjoyable. As with most new technologies, developers are learning on the fly and are struggling with very rudimentary SDKs/APIs exposed by bot platforms. In the meantime, we will have to put up with a lot of terrible bots (just like we have a lot of terrible apps) until we collectively learn the best UX/UI to engage users that uniquely suits bots.
2 — Platforms are still experimenting as well. This may be intuitive, but an interesting takeaway from the panel came from Rod McLeod of Kik: although Kik beat Facebook to the market with their bot store launch in April 2016 and have been experimenting with bots since 2014, Kik is still in experimentation mode as it figures out how best to engage developers and users. This experimentation includes what APIs to expose and how much control to give developers to test new use cases and ways of engaging users. This feels analogous to how the first mobile app stores emerged and how much control the platforms were willing to give developers, but hopefully platforms will learn from prior experiences and expose better tools sooner
3 — Chatbots are not a fit for EVERY experience, but they’re being built for EVERY experience. One of the larger contributors to the swath of “meh” chatbots is that developers are experimenting with bots across every possible use case. Experimentation is the right path for identifying the ideal use cases, but once developers/platforms narrow down on the key use cases, we will then hopefully see a higher density of quality chatbot experiences. The consensus from the panelists and Q&A participants was that chatbots currently seem like strong opportunities for games, customer service applications, parts of e-commerce and search. But even within those categories there are good and bad experiences. For example, I would definitely use a concierge chatbot service (especially fashion) given the low downside of a bad decision (free returns), but am unlikely to book a flight using the Hipmunk Bot on Messenger, because I would be paranoid at seeing limited options for a $500 flight and making a poor decision. I had the chance to speak with Hipmunk founder Adam Goldstein at the recent “Chatbots & Intelligent Agents” event hosted by Orange.
Adam shared that the true goal of the Hipmunk bot is not to book flights, but to capture travellers earlier in the funnel so that when they’re ready to book (typically after 15+ searches), they book through Hipmunk on mobile or desktop. The Hipmunk example illustrates the point that the more a chatbot can be utilized as a supplement rather than a substitute, or be incorporated to facilitate low-downside decisions, the more likely we are to see higher user engagement.
4 — We have yet to see a strong indication that bots as a business can stand alone. As the panel moderator, I was most interested in hearing the perspective on the question “can bot-first startups be stand-alone businesses, or will bots only be incorporated into existing startups?”.
Although everyone was bullish that bot-first startups (startups that are not leveraging pre-existing owned content, but rather provide value exclusively via the bot/chat-bot feature) will eventually succeed as stand-alone businesses, the vast majority of current successful examples of chatbot implementations are those that are leveraging pre-existing content, like the Funny or Die bot on Kik. The most promising bots to date seem to be in the enterprise space, mostly on the Slack platform, that are integrating various currently siloed stacks on Slack, like your CRM and internal feedback systems. Bots like Sudo have raised funding and demonstrate potential to charge for their integration services. However, it’s clear that the primary successful usage of bots in the near future will be as a new channel for existing content (like the CNN Facebook Messenger bot) rather than as a new business in itself.
5 — Platform risk is real, but messaging app usage will stay diversified. The second highest question on my list, and the one that garnered the most Q&A from the audience, was how developers should navigate platform risk. To access real scale and distribution all third party-developer chatbots currently can sit on one or a few of a limited number of platforms, such as Messenger, Slack, Telegram and Kik. Messenger and its billion+ users is the current gold standard for consumer bots, while Slack and its highly-engaged 3M+ DAUs offers the best opportunity for enterprise bots. But the allure of those platforms comes hand in hand with a dependency on these platforms to not change terms, APIs, access. For example, many bot developers are nervous that Facebook will follow the same path it did for brand pages where it encouraged the building of followers/likes and then basically required the page-owners to advertise to access their users. And why wouldn’t it?
Amit Garg had great insight on this risk sighting a statistic that the average user of a messaging client uses 3+, which seems to be growing rather than consolidating, as messengers specialize in different areas (Snapchat for ephemeral, WhatsApp/WeChat/Line for global communication, Kik for teens, Facebook for accessing all, Slack for enterprise, etc.). This diversity of messengers and use cases can help create a pressure on platforms to be developer-friendly in a way that Facebook never experienced once it emerged as the top social platform.
6 — We’re an Oculus or Pokemon-Go away from investor validation. While we’re seeing a ton of interest in the bot space, investment is still rather light, mostly in smaller angel rounds or through the Slack Fund. In my humble opinion, we’re an Oculus or Instagram moment away from seeing real investment. Both Oculus (for VR) and Instagram (for social/photo) were landmark acquisitions (both my Facebook) that proved the value of these unproven markets and demonstrated a path to liquidity for investors. While VR will certainly be more transformational as a technology than bots, a Facebook acquisition of a specific bot in the $20M+ range would help bot developers access the checkbooks of VCs that are currently sitting on the sideline waiting for proof points.
* NOTE: After launching the panel, we found a great article “To Bot or Not to Bot” by Amir Shevat, who is a friend of the Accelerator, an incredible thought leader in the bot space and runs Developer Relations for Slack. While the similar naming is coincidental (great minds!), Amir’s article is a great read and recommended reading for all bot enthusiasts.
Gary Coover is a tech and startup business model junkie who honed his snark through years of strategy/BD work, co-founding a startup, and a few years in Korea and the Bay Area working for Samsung. Gary currently runs Global Operations for the Samsung Accelerator, helping architect, launch and scale the Accelerator and its startups in New York, San Francisco and Tel Aviv. His opinions are his own, as are his tweets, which are occasionally above average.
Additional content from this author:
- What Entrepreneurs Can Learn from Olympic Silver Medalists — The Importance of Market to Your Startup
- Let’s Stop Throwing Out the On-Demand Baby with the On-Demand Bathwater — Four tips for validating strong on-demand business models
- Design as a Differentiator is Dead — Long Live the Business Model!!!
- Finding the Right Incubator “Unfair Advantage” — In the end, the unfair advantage is the key differentiator for both start-ups and incubators
- Frame Your Next Project for Success: a 6 Question Cheat Sheet — A former consultant talks about something other than hotel points and frequent flier miles
- The Startup Revenue Model Smell Test — Why Path may be in trouble
- The Lasting Tech Innovation of our Time: Putting a Price on Talent — Through the use of acqui-hires, Facebook, Twitter and Google have led the seismic shift in the way teams and talent are valued
