01
Focus
Direct attention toward the relevant mechanisms and questions.
New field guide · 28 pages
Thinking Patterns for Generative AI
Design better reasoning, not just better prompts.
A practical system for applying mental models, diagnostic lenses and structured critique to AI-assisted work—without treating frameworks as magic words or AI output as truth.
Built for: strategists, product leaders, transformation teams, consultants, and anyone using AI to investigate or decide—not only to draft.
14
Sieve stages
12
core patterns
3
worked cases
What changes
FIG. 01Most prompts specify the destination. Strong prompts also shape the route.
02
Decompose
Break complex work into visible, inspectable reasoning moves.
03
Evaluate
Define what a useful answer must reveal, challenge, and compare.
04
Verify
Separate facts, assumptions, hypotheses, and recommendations.
Inside the guide
FIG. 02- 01
From output requests to reasoning architecture
Why fluent, plausible answers still plateau—and what a thinking pattern changes.
- 02
How patterns steer an AI response
Attention, decomposition, evaluation criteria, contrast, and the limits of prompting.
- 03
Choose the lightest useful mode
Quick Lens, Structured Analysis, Full Constructive Sieve, or Critical Review.
- 04
The Constructive and Critical Sieves
Fourteen stages for framing, diagnosing, deciding, stress-testing, and communicating.
- 05
Pattern families and combinations
Practical routes for root cause, adoption, product discovery, strategy, and risk.
- 06
Three controlled before/after cases
The same context, a different reasoning architecture, and a transparent scoring rubric.
- 07
Failure modes, validation, and templates
Where pattern prompting breaks down and how to verify before acting.
A before / after moment
CASE 01Output-only prompt
“Analyse why our teams are not adopting AI.”
Likely result: a broad list spanning training, culture, leadership, governance and tools—with little help deciding where to intervene first.
Pattern-guided prompt
“Diagnose the adoption problem using Jobs To Be Done, Systems Thinking, Theory of Constraints and a pre-mortem. Show what each lens reveals and hides, then identify the smallest test.”
Likely result: a structured diagnosis of user progress, feedback loops, the limiting constraint, failure paths, and a testable next move.
The guide is careful about the distinction: better structure is not guaranteed truth. Pattern-guided work still needs evidence, external feedback and human judgment.
About the author
Raphael Thys helps teams turn AI adoption into an operating practice.
The Thinking Patterns Sieve grew from facilitation, AI training and strategy work: a way to make the reasoning method visible, choose only the depth a decision needs, and challenge an answer before it becomes action.