If you’ve spent any time looking at operational efficiency in our industry lately, you’ve probably discussed “automation.” Everyone wants to automate everything.
From dispatching in the field to back-office invoicing, the promise is always the same, push a button, save time and money.
But out here in the real world, we know it’s rarely that simple. Automation is an incredible tool, but deploying it blindly is a recipe for disaster. Let’s break down the realities of our digital workflows into three categories: The Good, The Bad, and The Automated.
The Good: Where Automation Wins
Keep it Simple
Automation is brilliant for rule-based, repetitive, and high-volume tasks. Things like manual data entry are mind-numbing for your team but perfect for machines. Automating these eliminates bottlenecks, reduces manual errors, and frees us up to focus on actual strategic work. The rule of thumb is, if the job is to do the same thing 10,000 times, automation is your best friend. If it requires doing it differently every time, you need a better process.
Don’t Reinvent the Wheel
Whenever possible, use platforms where templates already exist—like Zapier, N8N, or Power Automate. Just because you can, don’t write hundreds of lines of code to move an email to a spread sheet. Leveraging existing templates and integration tools significantly reduces implementation time. Why do you spend days building a custom pipeline when you can use a pre-built connector to start seeing immediate value?
The Bad: What to Avoid
Not the Hill to Die On
Don’t automate tasks that are only done periodically or take seconds to perform manually. Look at this classic XKCD Is It Worth the Time chart, it illustrates how spending 20 hours to automate a task that only takes 5 minutes a year is a terrible return on investment. You must weigh the setup and maintenance costs against the actual human time saved.
First Fix the Process
As Bill Gates famously noted, “automation applied to an inefficient operation will magnify the inefficiency”. You must streamline, standardize, and fully understand a workflow manually before you automate it. Automating a broken process just means errors now happen at machine speed – congratulations, the error has been successfully repeated many times!
Ignoring the Human Element
Avoid automating tasks that require complex human judgment, creativity, or relationship building. For example, you can automate client relationships if you want 100 perfectly written emails sent to the same person. Machines cannot handle gray areas well. Instead of trying to automate human interactions, automate the painful data-gathering tasks around your team to help them make better decisions faster.
The Automated: Navigating LLMs, Systems, and Tech Debt
Don’t Fight the Giants
Don’t spend time building custom automations where LLMs are already being implemented. Trying to out-engineer massive tech companies on native application features (like Outlook with Copilot) is a waste of resources. By the time you build it, Copilot or Claude might have it built-in. This means you are outgunned before you start. Focus your efforts on your unique, proprietary business logic instead.
Keep a Human in the Loop
LLMs can help with things like fuzzy name matches in databases, reading PDF scans of field tickets, or dynamically adjusting code. While LLMs can also be used to handle unstructured data, keeping a human in the loop is essential for catching unexpected variables. Fully autonomous processes can cause major security or operational headaches when they encounter edge cases.
Beware of Technical Debt
Every time you build a customized automation script instead of using a standard software feature, you are taking out a loan. That loan must be paid back in endless maintenance hours whenever an API changes or software updates. Keep it simple. Don’t build a fragile web of automations that will collapse the second the person who built it goes on vacation.
Know the Difference
Finally, don’t mistake system configuration, data validation, or reporting for process automation. Many teams waste months building complex “automations” for things that could be solved by simply checking a box in their existing software settings or building a data warehouse. Running the same SQL Agent job on Monday is not automation. This is like calling microwave leftovers “cooking”. It might work but it is also not automating, it is reporting.