There is a phrase I have heard throughout my career: “We need to automate this.”
Sometimes we do. But before we start talking about APIs, AI, workflow engines, integrations, databases, or writing a single line of code, I think there is a much more important question: Should we be doing this process at all?
Because automating a bad process doesn't make it a good process. It just makes the bad process run faster.
Technology Shouldn't Be the First Question
One of the mistakes I've seen organizations make repeatedly is starting with the technology. Someone identifies a manual process and immediately the conversation becomes: Can we automate this?
My first questions are usually different: Why are we doing this? What decision is the person actually making? What information do they need to make that decision? What happens before and after that decision? And perhaps most importantly, does this process still make sense?
Processes have a funny way of surviving long after the reason they were created has disappeared. Someone created a spreadsheet ten years ago. Someone else added another approval. A new regulation added another step. A manager wanted another report. A system couldn't do something, so employees invented a workaround.
Years later, nobody remembers exactly why the process works the way it does. They just know: “That's how we've always done it.”
Then somebody comes along and says: “Let's automate it.”
That is exactly when we should be careful.
A Lesson From Claims Assignment
Earlier in my career, I worked with claims-assignment processes. On the surface, the problem seemed straightforward. A claim came into the organization and someone needed to determine who should receive it.
There were rules, territories, workloads, different types of claims, and people making decisions based on all of those factors.
The easy technology answer would have been to simply reproduce what those people were already doing in software. But that wasn't really the problem we needed to solve.
The important part was understanding how the decision was actually being made. What information mattered? Which rules were real requirements? Which were simply habits? Which exceptions actually mattered? What judgment were people applying?
Once you understand those things, something interesting happens. You stop thinking about how to automate the existing steps and start thinking about how to achieve the outcome.
That's a very different problem.
Minutes, Hours...Then Seconds
The existing assignment process could take minutes of someone's actual working time, but the elapsed time could be much longer.
A claim might arrive and wait for someone to review it. That person might need additional information. They had other work ahead of it. They might be at lunch, in a meeting, helping another customer, or gone for the day.
A process requiring only a few minutes of human effort could therefore take considerably longer before the assignment actually occurred.
When we redesigned that process around the decision instead of simply reproducing the manual workflow, the assignment could happen in seconds.
That distinction is important. We didn't make someone click through the same process faster. We removed the need for most of the process.
That is where the real value of automation comes from.
Don't Automate the Clicks
This becomes even more important today because automation is becoming incredibly easy. With AI, low-code platforms, APIs, robotic process automation, and modern SaaS tools, organizations can automate processes faster than ever before.
That's wonderful. It is also dangerous, because we can now automate bad ideas faster than ever before.
Imagine an employee receives an email, downloads an attachment, opens a spreadsheet, looks up information in another system, copies several values into the spreadsheet, determines a category, emails the spreadsheet to another person, and then that person enters the information into yet another system.
You could absolutely automate all of those steps.
But perhaps the better question is: Why are there so many steps in the first place?
Maybe the information already exists. Maybe the spreadsheet shouldn't exist. Maybe the second person doesn't need to be involved. Maybe the category can be determined when the original transaction occurs. Maybe the systems should communicate directly.
Maybe most of the process can simply disappear.
That is process improvement. Automation comes afterward.
Talk to the People Doing the Work
There is another lesson here that technology teams sometimes forget: The people doing the work usually know where the problems are.
They know which screens waste their time. They know which approvals accomplish nothing. They know which spreadsheet exists because two systems don't communicate. They know which information they enter twice. They know which rules don't make sense in the real world.
If you walk into that environment with a predetermined technology solution, you will probably miss much of that knowledge.
Instead, sit with them. Watch the process. Ask why. Then ask why again.
You may discover that what management believes the process looks like and what actually happens every day are two very different things. That difference is often where the biggest opportunities live.
AI Makes This More Important, Not Less
Today, every organization is asking where AI fits. That's a reasonable question, but I think we should resist the temptation to sprinkle AI across every existing business process simply because we can.
Before asking “How can AI automate this?”, ask “What are we actually trying to accomplish?”
Then ask: “What is the simplest way to accomplish it?”
Sometimes the answer will be AI. Sometimes it will be traditional automation. Sometimes it will be an integration between two systems. Sometimes it will be changing a business rule.
And occasionally, the best technology solution will be: Delete the process entirely.
Automation Is a Multiplier
This is the principle I keep coming back to: Automation is a multiplier.
Give automation a well-designed process and it can multiply productivity, consistency, speed, and scale. Give automation a poorly designed process and it can multiply complexity, mistakes, technical debt, and cost.
So before building the workflow, buying another platform, creating another integration, or telling the AI to automate everything, spend some time understanding the work.
Talk to the people doing it. Understand the decisions being made. Challenge the assumptions behind the process. Remove the steps that don't need to exist. Simplify what remains.
Then automate it.
Because the goal isn't to automate more work.
The goal is to need less work in the first place.
Thanks,
Michael Cronin
Website: https://www.michaelcronin.info
LinkedIn: https://www.linkedin.com/in/michaeltcronin/details/experience/
