Working with automation and AI engineering, it's possible to observe recurring patterns. When we see a situation occur more than 50 times, it's no longer a coincidence, but rather a behavioral market pattern.
In this article, I share an "honest conversation" about four fundamental mistakes that consultants, companies, and implementers make when trying to automate processes, and I discuss practical solutions to avoid them.
Error 1: Automating a Broken Process
The first mistake starts at the base. For a automation For this to work, the original process needs to be balanced. The most dangerous mistake is automating something that doesn't work well, because automation only... amplifies the defects in the process.
The Real Case: The Retail Company
A retail company approached us wanting to automate order management to gain agility. Upon analyzing the process ("getting our hands dirty"), we identified several obsolete processes. Due to employee turnover, many followed established procedures without understanding why, simply clinging to the "it's always been done this way" mentality.
The result was a process with 10 stepsincluding duplicate order entries and unnecessary or expired approvals.
The Solution: Refactor before Automating
You can't build a house without a plan. The solution was to stop everything, interview the parties involved, draw up the plans, and question the workflow.
- Before: 10 manual and redundant steps.
- After: 3 automated steps.
When refactoring (redesigning) the process before By applying technology, we reduce errors, costs, and frustration. As I often say: "Sometimes we're using a jackhammer to cut a cake." The problem isn't the tool (the jackhammer is powerful), but rather where it's being used. Complex tools don't save bad processes.
Error 2: Implementing without Cultural Change
This error is often overlooked, but it's where many automations silently fail. Tools are just tools; people operate them. A successful implementation depends on... 50% the tool and 50% the people.
The Phenomenon of Resistance
Often, a new technology enters the sector "unannounced." Someone simply arrives and says, "From now on we'll be using this." This ignores the fact that the employee is the expert in that job and that the change impacts their routine and way of thinking.
If the implementation is careless, a classic problem occurs: after a while, the operator Go back to using the old spreadsheet.Why? Because in a spreadsheet, he has control, he knows where the cells are, and he feels secure. The new tool, without proper acculturation, generates fear and a feeling of loss of control.
The Solution: Prepare the People
The solution involves workshops, training, and, most importantly, active listening.
- Make the employee feel like they are part of the process.
- Validate his experience.
- Show how the change benefits the quality of work life and the safety of the operation.
Cultural change is inevitable for quality. When employees understand the "why," technology enhances human talent instead of competing with it.
Error 3: Seeking 100% Automation
There is an illusory pursuit of perfection, where people believe that tools will do absolutely everything, eliminating human need. This leads to paralysis.
The Real Case: The Digital Marketing Agency
A company became eight months stuck trying to automate the last 20% from a reporting project. The system already processed data from various sources, but there were details external to the data—inferences, judgments, and exceptions—that the machine could not perfectly capture.
The Solution: Pareto's Principle (80/20)
The recommendation is simple: Automate the repetitive 80% and leave the final 20% to humans. These 20% generally involve eminently human characteristics: creativity, judgment, handling exceptions, and customer perception ("game reading").
Accepting that automation doesn't need to be 100% creates synergy:
- The Machine: It does what it does well (repetition, volume, processing).
- The Human: Do what you do well (analysis, decision, exception).
Error 4: Not Measuring Results
It may sound cliché, but it's still a recurring and fatal mistake. Often, the solution "disappears" or is cut because no one has been able to prove its tangible value.
The Real Case: The CFO's "It Didn't Work"
In an RPA (Robotic Process Automation) implementation for reconciliation—a task no one likes to do manually—the project was technically a success. However, some time later, when budget cuts occurred, the CFO said, "This didn't work, let's cut it."
Why did he say that? Because Nobody measured the previous scenario..
- How long did it take before?
- How many mistakes happened?
- What was the cost per transaction?
Without the "AS-IS" (how it is) scenario, it's not possible to compare it with the "TO-BE" (how it turned out) scenario.
The Solution: Define Metrics First, Measure Later
It's necessary to define key performance indicators (KPIs) before starting. Measure the time saved, the reduction in errors, and the cost. Even if the change generates initial discomfort (learning curve), the data is the corrective and defensive factor for the project.
If you don't measure it, you don't communicate value. And if you don't communicate value, the tool is only seen as a cost.
Conclusion
These four points summarize the difference between automation that fails and automation that transforms the business:
- Don't refactor before automating. (and you'll only amplify the problem).
- Prepare the people (The tool is only useful if someone is able and willing to use it).
- Aim for the 80/20 (Don't waste resources striving for perfection; leave the noble part to humans).
- Measure everything (If you don't measure it, you can't defend the solution, and it disappears).
Automation is powerful, but it's not magic. It requires competent people, business acumen, and the humility to understand that perfection is the enemy of good.
