First, do it by hand
I’ve seen dozens, maybe hundreds, of automations go awry because people didn’t know what they were automating.
Automation is meant to offload repeatable work to computers. It’s not meant to offload creative work.
So, how do you define creative work?
If a task is just being created, it’s not creative work.
If a task is still being iterated upon for efficiencies, improvements, or enhancements to the user/customer experience, it’s creative work.
If a task can’t be written down either because there’s too much variability (“well, if X happens, do Y.”), it’s creative work. An easier way to think about this is to ask the following question.
“Would training a new hire on this task take more or less than 2 weeks?”
If it would take you more than two weeks, including answering questions, oftentimes because of weird corner-cases, “1-off” changes, or changing requirements either from internal leadership or external stakeholders, then it’s creative work.
These are tasks worth doing by hand.((There are exceptions to every rule, but these rules encompass over 90% of cases, in my experience.))
Each of these tasks have a commonality of “enough.” Either they’re not good enough, they’re not repeatable enough, they’re not actually business processes yet, or they aren’t standardized enough.
How do you know when a task is ready to be automated?
When you’ve done it by hand 10-20 times, and not much has changed between the first and last time you did it.
Now, it’s ready to be automated.
Doing simple, time-sucking tasks doesn’t scale, but it allows you to scale. If you automate a task the right way the first time, you’ll never need to touch it again.
If you automate a task that isn’t ready, or you automate it the wrong way, now you’ve got one of two (or both) problems.
A computer you need to babysit
Bad automations that work at scale
I can think of several times in my life where I needed to clean up a mess created by my bad automation. This is not for clients (although that’s happened as well), this is for me.
I have an AI do something, then I change the process, but I forgot the AI was doing work. Now the AI is broken or obsolete, and, chances are, my new process is broken too.
I have a new process (say, automatically invoice a client, then put them in an automated email series, then send them a Google Form to do preliminary data collection, then, once they submit the form, analyze the data and determine which new email series to put them in) but the process doesn’t work well because it doesn’t reach our business objectives. Now I need to manually review which clients are at which stage in the automations, stop the automations, and manually recover.
This has happened most often with customer onboard flows. I’d try to use AI to onboard new customers, change a process, forget about the AI, and then end up looking foolish in front of new customers because I had an automation go awry. There was never anything harmful that happened, but we didn’t look professional. Now, we don’t use AI to onboard new customers.
This has happened with analytics tools. We’d change the way we’d collect, send, or store data, but forget about an automation in the middle, which would lead to bad data in our tools. This was always frustrating, because we’d need to hand-check the data to verify what was accurate, which takes a long time.
Oftentimes, fixing the errors caused by automations can be more time consuming than creating the automations themselves.
Automations for solid, unchanging processes have saved me hundreds of thousands, or millions, of dollars. Automations for solid, unchanging processes have saved, or made, our clients millions of dollars.
Automations for processes that weren’t ready yet have cost lots of money too.
Beyond the questions above, use the formula below to determine if a process should be automated.
Write down every step in the process
Do that process by hand and do not deviate from the process you wrote down.
If you can do that 10-20 times, and never need to deviate or say “it’s just for this one person,” you’re ready.
If you’re not ready yet, keep doing the work by hand. Sometime, you’ll get that process to the point where it’s “good enough.” Then, go back to step 1.