Most executives come to me with the same question: “Where should we be using AI?”

It’s the wrong question.

The right question is: “What’s actually happening in our processes right now?” Because until you can answer that, you’re not doing AI strategy; you’re doing AI theater. And AI theater is expensive.

Here’s the uncomfortable truth: AI doesn’t fix broken processes. It accelerates them. Every inefficiency you feed into an AI workflow comes out the other end faster, at higher volume, and with more confidence.

The unlock isn’t finding the right tool. It’s doing the diagnostic work first.

The Move Nobody Makes

The instinct is to jump straight to automation. What can AI do for us? So companies buy tools, launch pilots, and call it transformation. Six months later, they’ve automated their chaos and wonder why nothing changed.

What they skipped:

• What shouldn’t exist at all? Half the steps in most workflows exist because someone, somewhere, added them to fix a problem that no longer exists.
• Where are decisions being made poorly or repeatedly? The same judgment call being made by five different people is a process failure, not a people failure.
• Where is data being recreated instead of used? If someone is copying information from one spreadsheet to another, that’s not work. That’s a symptom.

AI can’t answer these questions for you. That’s the part that requires human inspection, which is exactly why most organizations skip it.

The Framework: Process Mapping with AI Intent

Pick one workflow. One core business process your team runs repeatedly. Weekly reporting. Lead qualification. Order fulfillment. Client onboarding. Pick one and write it out step by step, every step, no skipping.

Then tag each step with one of three letters:

E — Eliminate. This step shouldn’t exist. It’s a workaround, a legacy artifact, or pure redundancy.

A — Automate. Rule-based, repeatable, structured input/output. No real judgment required.

D — Delegate. Needs a human, but doesn’t need you.

That’s the base layer. Most process improvement frameworks stop here.

Don’t stop here.

The Two Layers That Reveal Everything

For each step, add two more dimensions:

Decision Type

• None — Pure execution. Someone is just moving something from A to B.
• Rule-based — The decision follows a clear if/then logic. These are prime automation candidates.
• Judgment-based — Context-dependent, nuanced, or relationship-sensitive. These require a human, the question is which human and at what point.

Data State

• Where does the data for this step come from?
• Is it clean, structured, and accessible — or does someone have to hunt for it?
• Is it being recreated manually when it already exists somewhere else?

The Data State column is where most organizations find their real problems. Not in the complexity of their decisions, but in the fragmentation of their information. Data living in five places. Manual reconciliation masking broken integrations. People acting as human APIs, copying, reformatting, forwarding, because systems never talked to each other.

What the Map Reveals

When you do this exercise properly across a real workflow, four patterns emerge:

Elimination opportunities are the highest ROI. Every step you eliminate is leverage without complexity. Steps like “copy data from Sheet A to Sheet B” or “reformat the weekly report for the leadership deck” aren’t work; they’re symptoms of a system that was never properly designed. Eliminate them before you automate anything.

Automation opportunities are real, but often misidentified. Most organizations reach for AI when they should reach for a simple workflow tool. Steps that are repetitive, structured, and judgment-free don’t need a large language model. They need n8n, Make, or a basic API connection. Reserve AI for the steps where context matters, where structured data isn’t enough, and language or reasoning is actually required.

Delegation opportunities expose founder/operator bottlenecks. If you’re doing data gathering, status updates, or routine follow-ups yourself, you’re not doing strategy; you’re filling gaps in your system. AI can assist here, but the real move is removing yourself from the loop entirely. The question isn’t “can AI do this?” It’s “why is this still on my plate?”

Hidden bottlenecks are where the real leverage is. These don’t show up as obvious inefficiencies. They show up as: decisions happening too late in the process because the right information didn’t arrive in time. Data living in multiple systems with no single source of truth. Judgment calls being escalated to senior people because no one defined the decision criteria. These bottlenecks compound quietly. Mapping them makes them visible.

Start Here

Take one workflow. Block an hour. Write it out step by step with brutal specificity — if you feel embarrassed by how manual it looks, that’s the signal you’re doing it right.

Tag every step: E, A, or D. Add the decision type. Note the data state.

What you’ll find won’t be a list of AI use cases. It’ll be a picture of where your organization actually operates versus where you think it does. That gap — between the assumed process and the real one — is where transformation starts.

AI is a multiplier. The question is what you’re multiplying.

Map the process first. Then decide what to amplify.