Choose work worth redesigning with AI
This path takes the foundational distinction between task automation and workflow transformation into application. Turn it into a one-page opportunity brief that supports a decision: which flow, based on what evidence, why now, and within which boundaries?
Applied output
An opportunity brief combines the selected workflow boundary, evidence of current performance, the mechanism creating friction, a human + AI redesign hypothesis, and success-risk signals that protect the experiment.
Five parts of an opportunity brief
The goal is not to sell an AI idea. It is to make whether the opportunity deserves investment open to inspection.
- Boundary statement
- Define the trigger, finish condition, outcome, and outcome owner in one sentence.
- Baseline evidence
- Show current time, quality, rework, or capability loss through at least two signals.
- Friction mechanism
- Explain how the queue, context loss, decision load, or rework forms—not merely its symptom.
- Redesign hypothesis
- State which change in human and AI roles should improve the outcome, and why.
- Success and guardrail
- Pair one success signal with a balancing signal that protects quality, risk, or human capability.
Starting case
Invoice capture got faster; the exception queue grew
Finance extracts invoice fields 70% faster with AI. Records with missing purchase-order numbers or price mismatches enter a shared exception queue, however. With no clear decision owner, payment time does not improve and supplier calls rise.
What is the AI opportunity here?
The opportunity is not extracting more fields. It is redesigning the flow from invoice receipt to correct payment. The brief should expose exception ownership, decision evidence, and payment outcomes separately from extraction speed.
Signals of a strong first workflow
Boundable
One start, finish, and accountable owner can be named.
Evidence-backed
The problem is visible through data, examples, or a recurring observation.
Safe to test
The pilot can be narrowed, stopped, and reversed when necessary.
Outcome-linked
Success is not reduced to AI usage or local task duration.
Common trap
“Choosing an AI tool first, then searching for a problem that fits it.”
Tool-first selection can create local speed while adding queues, quality loss, or hidden control work elsewhere. Start with outcome and flow, then assign human, AI, and system roles.
Check before the PracticeLab
Choose one flow from your work: what outcome does it create, where is the current friction, and what is the most important risk of a poor redesign?
Describe the flow from demand to meaningful outcome, not the name of one task.
Keep this
A strong AI opportunity is not the easiest automation; it is a workflow where outcomes can improve while risk stays visible and learning remains possible.