ChatOps & ChatBots

At Emerging Stack, we love clever things that make our lives easier. We found ChatOps to be an incredibly productive approach for collaboration.

ChatOps, for us, means there's a constant flow of information being shared, ideas being debated, tasks being assigned and lots of lolz.

We didn't plan to work this way, it just happened organically. By close-of-business on Day 1, we'd realised email was going to be way too slow. Innovation demands pace, quick-fire decision making, fluency & zero lag. Sentence-by-sentence, not page-by-page.

6 months later, here's our chat-stats (since Day 2);

• 54,218 chat messages sent
• 5 emails sent
• A phone call every few weeks
• Zero hours face-to-face
• $0 spent (all via free services or open source)

But there's more to ChatOps than just Chat. Where's the Ops?

With some tweaking, our chatroom became a heads-up-display of all the things! There's a constant flow of important information being logged;

• Every website link we've ever shared
• Information about wiki pages being edited
• Fully-automated CI/CD event notifications from our development, build and project management tools

Bigger software-dev teams also use use plug-ins for things like approving workflow items, merging code, running jobs, etc.

But we promised bots and the lolz too.

Meet Reg. Our first employee (he's actually volunteering). Reginal D Hubot II is our chat-bot/robot-butler/court-jester-bot. Regv1.0 could do some cool things like spam us with random images of pugs (a pug-bomb), check the status of our Amazon cloud servers and a few other tricks...he was like a dumb Siri.

2 months ago, we upgraded to Regv2.0, and he's a bit different. He does everything Regv1.0 does, but Regv2.0 has a personality...sort of. He's been logging every chat message for the last 2 months, storing them in his brain and now we can talk to him and he sort of knows what we mean - he basically mashes up everything he's ever seen using some basic AI.

So, when Mike's offline, sometimes I have some one-on-one time with Reg. Just check in and see how he's feeling (I'm that sort of boss)

Giant caveat - Regv2.0 knows everything, he has no concept of discretion and he's a bit unpredictable. So don't send a password to someone in a chat room, because Regv2.0 will likely spit it back out again one day. And like siri, my kids love chatting with Reg...but unlike Siri, I do worry he'll respond with something NSFW. Garbage in, Garbage out.

Fear not...Regv3.0 is coming soon with much smarter AI and hopefully a parental-filter.

So Emerging Stack is off-and-running, our prototype products are progressing through development (watch this space) and we would be months further back if it wasn't for our ChatOps approach. We're huge advocates for it, especially if you're starting something new and you're remote from your teammates.

How can you get started with ChatOps and your own robot-butler? We use a combination of technologies (docker, hipchat, hubot source, bunch of plugins, epichal, etc) but there's a simpler approach available now, which is to use Slack for your chat app - it comes with a built-in hubot integration...it won't be as clever as Reg, but you'll still get all the other value of integrating your other apps into a consolidated view and maybe some lolz too.

In the next few weeks, we'll upload the Regv2.0 code to Github once we've cleaned him up a bit. You'll be able to spin up a bot on your laptop (boot2docker) and have some fun with it in your own chatroom. And if you leave him running long enough, he'll end up with some sort of mash-up personality, just like our Reg.

The more bots the better (cheap labour and cheap laughs) - if you found this interesting, or you're already getting value out of a ChatOps approach, spread the word by sharing this post with your current/future chatops team.

This blog post is featured on our website.