Before Paving a Cow Path with AI, Redesign the Road

I’m seeing a pattern. A team gets excited that AI can finally automate a complex problem, they jump in and automate the hardest part of a workflow and ship it. This feels like progress, but leaves opportunity on the table.

Why? If a process exists, teams automatically assume it should exist. Most workflows evolved and solidified over time; they weren’t designed.

Manual data entry due to incompatible systems. Three-day approvals because of incomplete information. Twenty-field forms since the old paper one once required it.

AI removes old constraints, changing what’s possible. The question isn’t “How do we make this step faster?” It’s “Do we even need this step?”

Before automating anything, write down the desired outcome in one sentence. It needs to be a real business outcome, not a fake process outcome. For example, you don’t want to “transcribe an intake call faster.” Your real objective is to “get a patient scheduled quickly with verified insurance coverage.”

When you aim at the outcome, the place where automation creates value for your organization changes.

If you’re planning a redesign

Which steps exist only because information was hard to get? If AI retrieves it in seconds, remove the step.

Where are you automating something that just compensates for an upstream problem? Can you fix upstream instead?

What would break if this task disappeared? If the answer is “nothing,” remove it.

One final reality check: just because you’re using AI doesn’t mean you can skip the change management part. Redesigning workflows means people need to unlearn habits and trust something new. Plan for change management upfront or your automation will be a bumpy ride.