The appetite for automation arrives at exactly the wrong moment.
Companies don’t reach for automation when things are clear. Clear things are easy; nobody’s desperate to automate easy. Companies reach for automation when they’re drowning — when the volume has outrun the team, when the manual work is crushing, when everything feels chaotic. Which is to say: companies reach for automation at their moment of maximum ambiguity, and point it directly at the ambiguity.
Here’s the law this section exists to state:
You cannot automate ambiguity. You can only accelerate it.
—
Automation is a multiplier. It multiplies whatever it finds. Point it at a clear process — defined stages, agreed definitions, designed handoffs — and it multiplies clarity: the work moves faster, identically, reliably, at scale. Genuinely wonderful.
Point it at a process the business itself doesn’t fully understand, and it multiplies *that.* The follow-up sequence fires on lifecycle stages that were never defined, so customers get emails that don’t match where they actually are. The lead routing runs on a definition of “qualified” that drifted years ago, so the wrong reps get the wrong leads faster than any human could misassign them. The renewal reminder goes to the customer who churned in March, because the system that knows about the churn was never connected to the system sending the mail.
And here’s what makes automated ambiguity categorically worse than the manual kind: the manual version had a safety net you never noticed. Humans. The absorbers from Chapter 5, quietly catching the weirdness — *that doesn’t look right, let me check before I send it.* Every manual step was also an inspection point. Automation removes the labor, which was the point, and removes the inspection, which nobody priced. The wrong thing now happens at scale, instantly, politely, in your brand voice, to every customer simultaneously — and nobody notices for months, because the automation also removed the witnesses.
Manual ambiguity fails retail. Automated ambiguity fails wholesale.
—
I need to say the current-year version of this out loud, because the stakes just changed.
Right now, everyone is wiring AI into their operations — agents on the pipeline, models on the follow-up, intelligence layered onto workflows. I’m not against it; the last part of this manifesto says so at length. But understand what you’re doing when you hand an undefined process to a system that never hesitates: AI is the most powerful ambiguity accelerator ever built. It doesn’t just execute your unclear process faster — it *fills the gaps confidently*, inventing coherence your business never actually agreed on, at machine speed, with perfect grammar. Feed it drift and it returns fluency. The output reads like clarity. It’s ambiguity, wearing its best suit.
An organization that hasn’t defined “qualified” is about to have that definition made, silently, ten thousand times a day, by a system nobody can ask.
—
So the question before any automation project — software, workflow, AI, any of it — is not *can this be automated?* Almost anything can. The question is: **do we understand this process well enough that we’d be comfortable with it happening ten thousand times without a human looking?**
If yes — automate it, celebrate, take the afternoon.
If no, then the process needs something before it needs speed. It needs the unglamorous work: definitions, agreements, designed handoffs, reality. Which happens to be the exact sequence the next section is about — because most companies, offered that choice, still try to buy it in reverse.