Not sure if chat bot...


Chat apps are about a dime a dozen, many of which never reach the fame of things like AOL IM, Facebook or good ole IRC. When we first started building CIF (even before), we used to have these "IRC Bots", err apps.. that we could `/cif` for and get a quick response. It wasn't so much a lack of web interfaces or software development kits ('SDKs'), as much as it was- IRC was where we were! To "google" something meant leaving that sweet sweet IRC interface and typing something in, then, assuming we found something legible, pasting it back into the chatroom so we could talk about it with our friends... the conversation was happing around the data-points, not vs versa.


Today- chatbots are everywhere (i'm look'n at you Siri!), and while they're moderately useful in the house, we really don't like it when they're watching our every move, it's kinda creepy. Sure, it'd be nice if they could predict what we wanted, or provided us with subtle hints during our day to day- things we've asked it to remind us of (if i'm obviously looking for my car keys, JUST TELL ME WHERE I PUT THEM!).. but they don't and .. i'm not sure we really wanna get that far (iRobot?) just yet..


That said, with chatbots, we can hyper-focus those contexts and interactions to generate a more meaningful experience. If we're in a chatroom, talking about an indicator- the subtly of a bot PM'ing us and suggesting "hey! i know about that- here are some links.." mid conversation can be quite useful. You obviously don't want the bot to be too spammy, but with the right combination of query-ability and common sense, it can be the subtle difference between finding that breach you've been hunting for- and not.


Once you get past the creepy factor-

Screen Shot 2018-02-03 at 7.49.53 AM.png

Targeted bots... errr automations are quite useful. They just need to be used in the right context, not in your way so they're stepping all over you- but enough so you remember they're there. Not smart enough to give you answers you're almost never looking for (I don't give a crap how many Kardashians there are!), but common sense enough to figure out what you're trying to do and present that information in a way helpful and targeted.

In fact- Shopify presented a way in which they've basically re-architected their operations around some of these frameworks. It really demonstrated how, if used properly, these types of automations can really drive your operations process to the next level.


That's just operations- What about customer interaction?

You know that button on the bottom right hand of the page? The one that says "Chat WITH A REAL LIVE HUMAN!", while there are a TON of apps to help you do that, a lot of times we just want quick and dirty answers.. usually to things that are already in doc and either we can't find, or ... we're just lazy (or both!). The idea of hijacking an actual humans valuable cycles on something as silly as "how do i install this?", a lot of times is the difference between using your product and not. Not that the customer support people behind those chat's aren't great at their job, 9 times out of 10 it's actually a pretty decent experience. But 9 times out of 10, i just need a quick and dirty answer, like RTFM STUPID, but with a friendly tone- and a relevant link to the doc I was searching for.

As nerds and geeks get expensive- those support costs can add up really quick (believe me, I run a set of open-source projects.. I do the support myself- it's a lot of wor.. err fun :)). However, if your doc is even a little out of date, hard to read or even harder to find.. at best- you'll get an email asking for your help in googling something.. at worst- they'll just move onto a different project. Every 30s question costs you 15min in productivity (context switching is expensive). If you're a multi-person, well funded machine- you just rotate in and out the warm bodies and deal with the overhead costs (by hiring more sales people, pushing growth, getting sucked into that never ending cycle..). But what if you're not? How do you balance great, intuitive customer support, without all the context switching and overhead? Some of us don't want that overhead- not because we hate hiring folks- but those kinds of folks have a very low retention rate, which requires it's own type of overhead just to manage. Some businesses are geared for that... I ... am not. I'd rather spend time building stuff- leave the sales and support magic to other businesses that are better suited for that (I am a platform... build on me :)).

There's an app for that!

I've stumbled across a few promising projects in the past few months, the crew over at (a ZenDesk competitor) have a neat app called "Carin" (I think the ZenDesk crew have something too, I would imagine most help desks these days do- or some paid monkeys behind the scenes pretending to be robots... or both. hard to tell sometimes). Carin in and of her-it-self isn't anything special, but the integration and implementation is fascinating from a "product usability" standpoint. While the fact that "you just added a javascript search box to the lower right hand corner of your screen" isn't lost on me- the interactivity of which does leave a certain impression on the user. A simple search box has no- personality, which means, when i've completed the search I almost have no reason to remember your search box (or your product?). Give that glorified search box a name, a pseudo personality and some features- and all of a sudden our brain puts a little pin in it as something more meaningful. It somehow adds to the customer experience.. we know it's fake- we know at any point we could ask it a stupid question and get a "I'm not sure you want to do that ... Dave.". The lure of what it could tell us in the future, somehow drags us back in. It's ... Sticky.

There are TONs of other examples, all ranging from Siri, Rita, Carin, Alfred, Cortana, Alexa (and LEX via AWS) to, which as of this writing i'm almost positive that since it's free- is being used to train the real robots (eg: the ones in iRobot) how to, well.. ya know.. "Bender" us. Most of these frameworks are trying to solve both the "intent" and "personality" problems (eg: how do I figure out what you're NOT asking, and fake the rest..). Ever play 20 questions? If I can narrow down what you're NOT asking, I have a higher probability of figuring out what you are. Most humans try to think about this the other way around- decipher what you're asking- then give you an answer. As a bot, I only have a small universe of know-how to pick from, if you're not in my universe, the answer is going to be "i dunno". That's easy, so based on my universe if I can eliminate the other 84% of things that are lower probability, I only have to "flip a coin" on the last 16% of the higher probability stuff- yadda yadda yadda (it's just basic math, and knowing how to word things to fake the rest..). Most humans won't notice the difference, they just wanna be lead in the right direction, and the ones that have a hard time with it- they're probably too much overhead anyway. They're the ones that probably need dedicated support or a dedicated sales team- something that's a bit less interesting. The robots almost naturally help you weed those low margin folks out.

Yesterday- it was IRC.

Yesterday it was IRC (or Yammer- remember Yammer?), Today it's Slack, tomorrow... who knows. The tools and frameworks come and go, but the pattern still remains pretty constant. We use these different automations to help build interaction around our operations as well as our customer support. We give them quirky names and sometimes a personality to fake ourselves into believing they're more than a bot- It's almost self serving because we want to remember they're there. We want to believe that we're not just talking to a python script and we want our experience with them to be more productive. We don't always want to open a browser, or a tool or a widget to figure something out. It breaks the conversation, and as any human knows.. we get distracted very 




... what was I saying?

Right, chatbots... we're just getting started.

Did you learn something new?