Many companies rush into automation. Management thinks that by introducing auto-function, they will save money quickly. They want to run automated systems immediately that combine too many variations, too many human decisions, and too many disconnected systems. It is tempting to automate as much as possible. Think of a sci-fi movie you like–automation looks cool, right?
Not so fast.
The Webvan story from 2000 is considered the baseline for failed automation. Webvan decided to build its complex automated infrastructure from scratch—and all at once. Hundreds of engineers built the software algorithms to make five miles of conveyor belts that included transferring product into a tote and getting the product to the shipping dock. They also automated real-time inventory management to ensure that items were currently in stock as necessary, routed delivery vans to multiple stops, and dealt with real-time delivery confirmation or returns.Sounds awesome. The problem was that the automation was never fully developed correctly and never quite worked as well as intended. There were too many moving parts all at once. It was a disaster and Webvan went under.
Older automations started with the manual version and then developed from there. Take this Windward experience, for example:
One of my first gigs was at a University, where I was responsible for running the mail system for the administration. It was a painful system and part of what I had to do to work it out. Not only did I have to get it up and running, I had to make sure the system was totally managed. First I had to make sure the system actually worked. Then I figured out how to make the system run better and how to measure it: If things were good and things were bad. First I observed manually and then I figured out how to create automated processes to make the system self-monitor and self-diagnose its own problems. I made it more self-managed, more self-contained. And then, ultimately, I got it to run itself and tell someone when it’s got a problem and then fix itself. That’s what I do as part of the system lifecycle—fixing a system so it’s not the worst part of your life.
It is true, automation systems have matured since Webvan. IT doesn’t have to go to ground zero anymore or start from manual, because due to maturity in IT standardization and virtualization, and moving to the cloud, a lot of underlying components and infrastructure have developed and become standardized. Companies can start with the pre-organized parts and there are now pre-built monitoring services and instrumentation. You can leverage subsystems and facilities that are already there and they are straightforward.
So, yes, take advantage of new systems and advances in infrastructure. Leverage the heck out of subsystems and the cloud where you can. Just start slowly and make sure you have a defined automation plan in place that stakeholders follow. And make sure the first step functions properly before you go on to subsequent steps. And that each function after that works properly. You can’t automate everything at once. That will result in headaches, not cost savings. Don’t be the next Webvan.