The Four Observations for Software-based Automation (18/30)
I'm writing 30 posts in 30 days. This is number 18.
a16z has a piece about operational analytics and I like how it distinguishes digitization and automation. Most literature out there writing about automation tends to conflate the two. I will briefly explain why separating the two is important and how that helps automation occur faster and more successfully.
When Automation and Digitization are Different
The kind of software automation to reduce busy work relies on digitization to happen first. Digitization is simply turning manual processes into digital processes. This tends to get confused with automation per se. And understandably so, because sometimes — note I say sometimes — when you digitize a process, you would also have made it automated. Imagine if you switch from manual marking of multiple choice questions answers on paper to using typical OCR paper shaded with 2B pencils, you would have digitize and automate in a single move.
The questions are marked automatically by the OCR tool (hence automation) and the results are generated and stored digitally (hence digitization). Automation and digitization are conflated when the automation is on the input part of the process.
On the other hand, a lot of technology upgrade tends to be digitization alone. Imagine you replace a paper and pen process for warehouse workers tracking inventory to a digital web application. In both cases, the input part is still manually triggered — workers have to key in the data. However, the first case is definitely non-digital but the second the data after manual key in it becomes digitally stored and manipulated. This example is purely digitization.
If the input task is not automated, then it’s purely digitization.
Digitize then Automate is Another Automation Pattern
Sometimes, the biggest returns for investment in automation is to first do digitization, then automation. This digitize-then-automate process usually still relies on humans to key in the data into digital systems. However, once the data is in digital form, usual manual tasks like reconciliation, data sanitization, now can be automated.
Automating repetitive, rule-based tasks on already digital data is the low hanging fruit of software automation.
Having the Right Distinctions
I've been doing automation work that largely involves the digitize-then-automate pattern for several years. And for the most part, I was also conflating both digitization and automation. This is my mistake.
Don’t repeat my mistake.
Similar to writing software, whenever you don’t make the right distinctions at the right abstraction level, you end up writing software that don’t quite satisfy your users even if it passes all the unit tests.
So see the truth clearly for what it is. To summarize, I have the following four observations:
Automation != Digitization
When you automate the input into a digital system, automation looks indistinguishable from digitization. This might even involve hardware as well.
When you digitize-then-automate, you rely on humans to manually key in data but then you automate the data processing tasks on digital data. This is the low-hanging fruit of software-based automation.
2 is much harder than 3.
I wish I knew this long before I wasted years of my working life.