@chrismessina really interested in seeing what mass consumer interaction with a bot looks and feels like. I've been generally skeptical of bots that "look like humans," because users may expect human-level natural language capabilities, only to be disappointed. Is that a cultural jump we can expect American consumers to make as readily as consumers have in other markets?

@nkbuduma one thing @_amanda_ did that was clever in the design was that she made it very clear that Kalani made the bot for her fans - this created a link to the real influencer in her fans minds but it was clear it was a bot. @_amanda_ 's tricky like that 😉. She's here if you want to ask her any questions directly.

@andymauro @_amanda_ that's clever indeed :) so with a combination of a little good old self-deprecating bot humor and an intelligently situation disclaimer you can make a hard coded decision tree less frustrating?

@nkbuduma @_amanda_ Kik really promotes fairly static decision tree type designs so we were somewhat locked into using suggested replies and a linear flow, but there is a fair amount of ability to go off track with typed input if you so like. Our trolling filter We definitely learned that Kik users do in fact type input to bots, and our CLU engine did a pretty good job handling a lot of the inputs (but 12 year old fans are a whole new level of tricky to understand - figuring out intentional spelling errors and messaging slang is the new beam forming microphone). Interestingly we are currently hanging at just under 1% of all messages sent were trolling messages to the bot - this is an important reason to have CLU in your bot, just to catch all the rotten things people will say to it :o

@andymauro @nkbuduma @_amanda_ I am curious about the differences in behavioral norms and expectations between generations. My hunch is that kids will be more forgiving and willing to suspend disbelief because of their active imaginations, whereas older demographics will frustrate more quickly with today's state of the art bots.

@jonbruner @andymauro any early data on what engagement in this use case looks like?

@nkbuduma @jonbruner Hey Nikhil, the Ad Age article has a bunch of data and there's some in my response to Esther's question, but beyond that I can tell you that as of 5m ago 48% of the people that talked to the bot had a conversion event (extremely high for any kind of marketing), 53x more people talked to the bot than to the real influencer (and that number keeps getting bigger over time since the IG posts fizzle out after 48 hours or so, the Bot just keeps on ticking), and 88% of the users say they either love or like the bot, and only 6% of users say they hate it (the other 6% are in the middle) which is an incredibly high sentiment rating for any kind of chatbot. We and the client are quite happy with engagement numbers, but more importantly see a path to drive sales on the back of this strong engagement

@andymauro @jonbruner Those numbers are ridiculously compelling... do you have a sense of what about the bot makes it so appealing to users? It seems like you've tapped into a level of reliability that has escaped most people trying to develop bots. Would love to talk offline sometime about the insights you've used to build compelling bot experiences -- I've been working on bringing similar tech into healthcare which is in dire need of some innovation :)

@nkbuduma happy to talk - andy at automat dot ai. A few things helped get us to that level, Firstly, using the right influencer, as chosen by our partners at theAmplify who use AI tech to match influencers to brands - our partnership with them was and continues to be a key component to this. I don't have much confidence in bot-only companies trying to build celebrity or influencer bots that don't have a world class influencer marketing company behind them like we do with theAmplify. Secondly, we've been doing this a long time and have learned a lot of lessons building prior generation tech like IVR and mobile Virtual Assistants ;) That said, this was really a trial, we've got far more powerful tech and bot use cases coming, but this sets the stage nicely!

Hay @esthercrawford! Kik was a really fun platform to build on. Pros are that the partner team over there is awesome, we got help on the tech, biz-dev, design, and PR front and they were always super responsive. The Bot Shop is also a big benefit as we saw the bulk of the traffic after the Instagram campaign's 48 hour window come from there - it was also good timing that they were just in the middle of launching their Fashion and Beauty section. Cons are that the platform is less full featured than other platforms and so things like carousels, cards with buttons, persistent menu, etc are not available. Also, we had problems with users getting rate limited because Kik's model of ensuring bots don't spam users is kinda broken (i.e., it's really common for a bot to send multiple messages in response to a single user message, but Kik expects the bot to maintain a near one to one limit which really inhibits design a lot). Overall the most surprising thing about Kik is how engaged bot users are if you get it right. We currently have 1279 reviews in the Bot Shop with an average of 4 stars -- this is one of the most reviewed bots in the Bot Shop after only 4 weeks! It's pretty clear that teens like bots, are willing to spend a fair amount of time with them, and will even take a moment to review them - I wasn't expecting this. So, all in all I'm pretty hopeful about building on Kik in the future.