Sufficient CertaintyDeciding
The Universal Decision-Making Method

You have a decision to make

The Walk is a structured conversation that helps you think it through, confront what’s uncertain, and leave with a decision you can defend.

Start a Walk Read the book →
Decision Record
Ref: DEC-2025-117
Deploying AI for Quality Inspection
6 Assumptions
Model accuracy will meet the 99.2% regulatory threshold within six months Critical
Production line staff will operate alongside the new system without disruption Critical
Hardware supplier can deliver and install within the shutdown window Important
9 Sufficient Certainty
Assessment: Proceed with secondary elements. The regulatory-accuracy assumption retains Critical character: high influence, moderate confidence. Parallel-run protocol, staged rollout across two lines, and monthly accuracy audits together convert uncertainty into managed exposure.
11 Monitoring
Accuracy: false-negative rate vs. 99.2% threshold, weekly. Adoption: operator override frequency, fortnightly. ROI: unit inspection cost vs. baseline, monthly.

This is what a walked decision looks like.

Decision Record
Ref: DEC-2025-117
Deploying AI for Quality Inspection
1 The Decision
Whether to replace manual visual inspection on Lines 3 and 4 with an AI-driven quality detection system, at an estimated capital cost of $2.1M, targeting full deployment by Q3 2026.
3 Context
Current defect-escape rate sits at 0.6%, above the 0.4% contractual threshold with two key automotive clients. Manual inspection throughput cannot scale to the planned 18% volume increase. Three competitor facilities have adopted similar systems in the past 24 months.
4 Options Considered
A. Full AI deployment across both lines simultaneously
B. Staged rollout: Line 3 first, Line 4 after six-month validation
C. Hybrid model: AI flags, human confirms (permanent)
D. Staged rollout with parallel-run period (selected)
5 Stakeholders
Operations Director (accountable), QA Manager, Line Supervisors (3 and 4), IT Infrastructure Lead, Procurement, two automotive client quality liaisons.
6 Assumptions
Model accuracy will meet the 99.2% regulatory threshold within six months Critical
Production line staff will operate alongside the new system without disruption Critical
Hardware supplier can deliver and install within the shutdown window Important
IT can provision GPU compute without impacting existing MES uptime Important
Training data from Lines 1 and 2 will transfer to Lines 3 and 4 product mix Important
7 Assumption Map
Influence
CRITICAL
Model accuracy will meet the 99.2% regulatory threshold
Staff will operate alongside the system without disruption
IMPORTANT
Hardware supplier delivers within the shutdown window
RELEVANT
GPU compute runs without MES impact
LIMITED
Training data transfers across production lines
Low Confidence High
9 Sufficient Certainty
Assessment: Proceed with secondary elements. The regulatory-accuracy assumption retains Critical character: high influence, moderate confidence. Parallel-run protocol, staged rollout across two lines, and monthly accuracy audits together convert uncertainty into managed exposure.
11 Monitoring
Accuracy: false-negative rate vs. 99.2% threshold, weekly. Adoption: operator override frequency, fortnightly. ROI: unit inspection cost vs. baseline, monthly.
Start your own Walk →

“You’re not likely to love risk management when you’re done with Deciding, however, and that’s all for the best. The book moves the conversation away from consultant-driven boardroom BS into a more practical, and infinitely more applicable, method that anyone can use.”
— Christopher Paris, Founder & VP Operations, Oxebridge
“The authors deserve accolades for rejecting jargon and writing in plain English with words having their normal meanings. Decisions are not ‘fire and forget’.”
— Warwick Stacey, Former special forces officer

Origin

The Universal Decision-Making Method is the work of Grant Purdy and the late Roger Estall — three decades of partnership across mining, finance, aviation and government, distilled and set down in Deciding in 2020, their last work together. This tool is built by Grant’s son, Jo, to keep the method in the hands of the people who need it: Deciders making one decision at a time, in plain English, without the millstone. It is not a software product wrapped around a book. It is the method, made conversational, with the book’s voice still in the room.


Deciding — A guide to even better decision-making, by Roger Estall and Grant Purdy

Deciding

By Roger Estall and Grant Purdy (2020). A practical guide to better decision making for leaders, set down in eight chapters and five appendices. 160 pages of the method, without jargon.

Available on Amazon
Have a decision on your mind? Start a Walk →