Can, or Should
The second axis of AI deployment
Deciding not just what AI can do, but what work to automate and what to protect.
Most AI asks the wrong question
Almost every automation effort runs on one question: can a machine do this task?
That is a single axis: logical feasibility. It produces two failures, everywhere.
Over-automation
We hand machines the work that should have stayed human.
Under-automation
We leave humans doing the drudgery no human should be doing.
Twenty years ago, I scored ideas for a living
First hire in the IP practice at Pangea3. Patent analytics and IP valuation at CPA Global, the largest IP management company on earth. I hold patents on scoring and classifying patents.
My job was to put a number on an idea and decide who should own it.
That is exactly why the second axis, and the decision to give it away, are deliberate.
The second axis: Cognition & Empathy
◆ Cognition
How complex, consequential, and expertise-bound the work is.
◆ Empathy
How emotional, relational, and personal the work is.
One axis tells you if a machine can. Together they tell you if it should.
Two axes, in detail
◆ Cognition 0–40
- Decision complexity — conditions, variables, branches
- Consequence severity — what breaks if it is wrong
- Context dependency — external information needed
- Expertise depth — how specialized the knowledge
◆ Empathy 0–40
- Emotional stakes — does it touch how someone feels
- Relationship impact — does it shape a relationship
- Individuation need — how personal each instance is
- Communication sensitivity — privacy, nuance, channel
The answer falls out of the grid
The thank-you that matters and the hard piece of feedback are not edge cases. They are the protected quadrant, kept human on purpose.
The scores drive a redesign
Each task maps to an Execution Authority Protocol: who acts, who oversees.
A pattern library restructures the flow.
The system outputs multiple future-state architectures, each optimized for a different goal.
A linear, human-paced process becomes an agent-native one, by design.
This is Reimagination
Not digitizing the old workflow faster. That is paving the cowpath: faster waste.
It is redesigning how the work is done so humans and agents share the load, and the human parts are protected on purpose, not by accident.
Why give it away
A decade in the IP world taught me what the field does not advertise: the most valuable ideas are not the ones you fence off. They are the ones that become how everyone works.
A defensive publication makes an idea permanently un-ownable. No one, including us, can patent it.
Choosing influence over rent. Standard-setting over toll-collecting.
Start here
Pick one workflow you are about to add AI to.
Score each task on cognition and empathy first.
Find the relationally sensitive quadrant and protect it.
Redesign around the scores. Do not pave the existing path.
Ask should we before can we.
Dual-Axis Scoring for Automated Synthesis of Operational Architectures
Co-authored with Hariprasad Rengarajan.