DevOps Cat’s Guide To Automation - Part Two

How to train your human operator to respond to incidents

Yoko Li · Jul 10, 2020

As I spend my pandemic days WFH thinking about how human-in-the-loop automation can make life easier for on-call engineers, Pixel and Cookie have been keeping me company. I realized one day that they have been automating around me for years, creating the perfect interactivity between human and machine to maximize their desired outcomes. They are the DevOps Cats, and this is their story. If you missed part one, read it here.

Part Two - Why are human operators important?

Kitten: "Why can't I just use AI and machine learning to automate incident management? Cat: "Well, kitten. I moved in with humans for a reason."

Humans ask better questions. Cat: "I'm not feeling well..." Human: "Where is it hurting? Did you get injured from fighting with the other cat? Did we forget to play with you so you are feeling depressed?" Robot: "Are you a cat? Please answer yes/no."

Humans can develop intuition from unrelated events with intrinsic similarities. Cat: "I don't know why the other cat always eats from bowl during dinner time. She has her own bowl." Human: "That reminds me of a load balancing problem. Let me try to direct her to another food source away from yours during dinner and try to debounce how many times we feed both of you." Robot: "What is the other cat's name? Variable not defined."

Humans have high emotional intelligence, which is important during a stressful incident. Cat: "I'm so scared of the fireworks that I accidentally peed on the carpet." Human: "That's okay! Let's think about things we could do to make you less scared next time to prevent incidents like this! Incidents are unavoidable, but we'll do our part to make it a little better to handle." Robot:"What attributes of the fireworks are you scared of?"

That's why I chose two humans as my minions instead of a machine

Now that Pixel and Cookie understand that humans should be a part of the automation process, will Pixel and Cookie teach their human how to automate more strategically? Stay tuned for our next installment, and follow us @Transposit.

Try intelligent runbooks and simplified incident resolution