The 93-Step Workflow Is Not the Flex

I used to be impressed by screenshots of huge workflows.

You have seen them. A Clay table with ninety columns. A canvas with so many nodes it needs a scroll bar. Caption reads "this took me three weeks."

I would look at it and think that person is better at this than I am.

Then I talked to a revenue ops lead who had built one of those. He said something I keep repeating.

"I have a 93-step Zapier app, and it sucks. That's annoying as hell to manage."

Then this: "It's funny that people want to show that they can handle complexity and that they're so clever, when if you have elegant enough design, nothing should have to be that complicated."

He built it. He can maintain it. He is still tired of it.

The two tired people

Two kinds of people look at that screenshot, and both feel bad.

The first one cannot build it. She sees ninety columns and assumes that is the bar. She opens her own table, gets four columns in, and stops. Not because she is slow. Because nobody ever showed her what finished is supposed to look like. I hear a version of this on almost every call. I do not know if I am on the right track, or if this is as good as it gets.

The second one can build it. He did build it. Now he owns it. Every small change means opening the whole thing and remembering what past-him was doing.

Nobody there is happy. And complexity keeps getting sold as proof of skill.

Why it happens

I do not think anyone is being dishonest. I think it is easier.

Complexity is what you get when you never go back. You add a column for a case. Then another for the case that broke. Then a branch for the thing you did not expect. Every one of those was the right call in the moment.

Nobody sits down and designs a 93-step workflow. It grows.

AI makes that growth much faster, and that part is worth saying out loud. AI is really good at scaling chaos. You can build the mess in an afternoon now instead of over a quarter.

What we found

The good builds we see are smaller than people expect.

Not because we cut corners. Because most of the extra steps were doing the same job twice, or doing a job that a filter earlier in the flow would have made unnecessary.

Three patterns keep showing up.

Filter before you enrich. Half the steps in a big build run on rows that get dropped later. Move a cheap filter to the front and a lot of the flow disappears on its own.

Use a rule where a rule will do. An AI column that scores a lead and also decides is doing two jobs. Let the AI score. Let an if/then decide. If the score is under 7, reject. Rules do not drift. Agents do.

One place to change one thing. If a fix means editing more than one table, that is not a system. Those are copies.

The honest version

Smaller is not automatically better.

Some work genuinely is complicated. A waterfall across five data providers has to be a waterfall. Some businesses have eight real segments and each one needs its own scoring. That is not bloat.

So the test is not step count. The test is this.

Can one other person open it and say what it does?

A big system a team can maintain beats a small one that hides what it is doing. And a small one that works beats a big one nobody dares touch.

I would rather hand back something a client can change without me. That is the flex.

I might be wrong. Maybe somewhere there is a business where all 93 steps are load bearing.

I have not met it yet.