What to Automate First, and What to Leave Alone
Most automation projects don't fail because the technology was wrong. They fail because the wrong work got automated. A process gets picked because it is visible, or because a vendor demo made it look easy — not because automating it would actually change anything.
Here is how we sort it out with clients, before anyone writes a workflow or buys a tool.
Automate the work that is repetitive, rule-based, and boring
Good automation candidates share a shape. The task happens often. The steps are the same every time. The decision points can be written down as rules a reasonable person would agree on. And nobody on your team would miss doing it.
Typical examples in small and mid-sized businesses:
Copying data between two systems that do not talk to each other
Sending the same confirmation, reminder, or follow-up message
Assembling a report someone rebuilds by hand every Monday
Routing an incoming request to the right person based on a few fields
Checking that a record is complete before it moves to the next step
These are the tasks that quietly consume hours. They rarely appear on anyone's job description, which is exactly why they stay invisible until you go looking for them.
Protect the work that carries judgment, trust, or relationship
Some work looks repetitive from the outside and is not. A customer call that follows the same script every time still depends on the person hearing what is not being said. An approval that is "always approved" is still the moment someone catches the one case that should not be.
Be careful automating:
Conversations where the customer is upset, confused, or deciding whether to stay
Judgment calls where the cost of being wrong is high and the signal is subtle
Steps that exist as a deliberate check on something else
Anything that is currently the reason someone chooses you over a competitor
Automating these does not just risk mistakes. It removes the part of the work that creates loyalty, and you usually do not find out until the numbers show it months later.
Four questions to ask before you automate anything
How often does this actually happen? Ask for a count, not an impression. "All the time" is often twice a week.
What happens when it goes wrong? If the failure is invisible, expensive, or lands on a customer, the check belongs in the design rather than bolted on later.
Who does this today, and what would they do instead? Automation that frees four hours and fills them with nothing is not a win yet.
Would we still want this process if it were instant and free? Sometimes the honest answer is no, and then the right move is to retire the process rather than automate it.
Fix the process before you automate it
Automation makes a process faster and more consistent. It does not make it better. If a workflow has three approval steps that exist because of a problem you solved two years ago, automating it means you now do the wrong thing at scale, reliably, forever.
The order that works is simplify, then standardize, then automate. Most of the value usually shows up in the first two steps, and they cost nothing but attention.
Measure the thing you actually care about
"We automated invoicing" is not a result. "Invoices go out the same day instead of Friday, and we are paid nine days sooner" is a result. Decide before you start which number should move, then check it eight weeks later. If nothing moved, you learned something useful and cheap.
Where this usually starts
In practice, the first honest conversation is worth more than the first tool. We spend time understanding how the work really runs today, including the workarounds nobody documented, before recommending anything. That is the beginning of the ETS Path, and it is why our approach starts with listening rather than software.
If one process in your business is eating a day a week, that is usually the place to start. Not the most exciting one. The one costing you the most quietly.
Not sure where AI or technology could help your business?
Tell us what is taking too much time, what is not working, or what you would like to improve.
