AI Automation · Foundations
Automating a bad process just makes the mistakes happen faster.
Not every repetitive task is ready for automation, and some never will be. Before spending money connecting two systems, it's worth checking the workflow against a short list of signs, because the workflows that fail after launch usually failed this check first.
Direct answer
Five signs it's ready
It happens often
Daily or weekly, not once a quarter. The time saved on a rare task rarely justifies the build.
The rules are consistent
The same inputs lead to the same decision almost every time, with a small, nameable list of exceptions.
It spans tools that don't talk to each other
Someone is retyping the same information from a form into a CRM into a calendar. That retyping is the automation.
The process itself has stopped changing
You've been doing it the same way for a few months, not redesigning it every other week.
Success and failure are both obvious
You can tell, without judgment calls, whether a given run worked. That's what makes error alerts possible.
The other side
Three signs it isn't ready yet
It needs judgment, not rules
A customer complaint, a pricing exception, a decision that depends on reading the situation. This belongs to a person, or to an AI draft a person reviews, not a fully automatic step.
The volume is too low
A task that happens a few times a month usually costs more to automate and maintain than it saves. Simplifying it by hand is often the better fix.
Nobody agrees on the "right" way yet
If three people on your team would each do it differently, automating locks in one version before the process is actually settled. Standardize first.
Why this matters
Automation is one of five possible fixes, not the default
A workflow that fails this check isn't a dead end. It usually means a different fix: writing the process down and standardizing it first, buying an off-the-shelf tool instead of building a custom connection, or handling it with an AI-drafted, human-reviewed step instead of a fully automatic one. Automation is the right call only when a workflow is frequent, rule-based, and stable enough to trust running without a person watching every step.
Where it fits
This is the first question in every automation project
Before we design or build anything, discovery starts by mapping the workflow and checking it against signs like these. If it's ready, we scope a one-page spec with the trigger, the fields, and the approval points before any build starts. If it's not, we say so, and recommend the fix that actually applies.
Questions
FAQ
What if a workflow is only missing one of the five signs?
It depends which one. A low-volume but otherwise perfect fit is usually fine to wait on. Rules that still shift week to week are worth fixing first, since automating an unsettled process locks in the wrong version.
Can part of a workflow be automated and part stay human?
Yes, and that's often the right shape. The data-moving steps get automated; the judgment call in the middle stays a person's, sometimes with an AI-drafted recommendation to review.
How do you check for these signs?
Discovery starts with watching how the workflow actually runs today, not how it's described. That's usually where the real frequency and the real exceptions show up.
Next step
Let's check your workflow against the list
A short call to walk through a process you're considering automating, and find out honestly whether it's ready.