From ambiguous mandate to a reviewable decision model.
A fictional worked example of the six-step decision workflow, followed by its outputs. The names, scores and simulation below illustrate how to inspect and question a model; they are not independently validated business results.
Methodology
AHP + DEMATEL
Illustrative simulation
1,000 Monte Carlo trials
Review focus
Assumptions and trade-offs
Shown using
Fictional business example
01 / 07
Context
Start with the decision your team needs to make.
"We have five initiatives, capital for two, and 26 months of runway. Which should we prioritise?"
Capture the goal, constraints and people affected by the decision. In this illustrative example, a fictional software company compares growth initiatives. AI can help surface questions and refine the brief, while the decision owner checks and saves the assumptions.
Illustrative brief — SpectrumOps
Fictional worked example
"We are SpectrumOps, a B2B SaaS company providing AI-powered supply chain visibility... closed our Series B nine months ago. The board expects a clear path to 10M euros ARR within 24 months.
Runway: 26 months at current burn. Aggressive hiring across multiple initiatives could compress this to 16–18 months.
Engineering team (28 people) is already at approximately 85% capacity on core product.
Initiative 'AIEdge': Develop advanced predictive capabilities... requires hiring 3–4 senior ML engineers in a highly competitive talent market. Estimated investment: 2.5M euros over 18 months..."
HUMAN CHECKPOINT
Confirm what is known and what is assumed. Runway and engineering capacity need explicit review before they guide the model.
02 / 07
Criteria
Agree on what success means.
Which factors matter for this decision, and which would change how we judge the options?
Draft criteria manually or with AI assistance. Inspect the set for missing concerns, overlapping factors and unclear rating directions. The example uses seven criteria, including execution feasibility. Accepting a draft records a choice; it does not establish the quality of that choice.
Illustrative evaluation criteria
7 example criteria
Execution Feasibility & Resource Availability
Constraint-Derived
Strategic Alignment & Market Positioning
Board Priority
Market Size, Growth Rate & Competitive Landscape
Financial Upside & Return on Investment
Risk Exposure & Mitigation Complexity
Time to Revenue & Implementation Horizon
Sustainable Competitive Advantage & Moat Creation
REVIEW BEFORE ACCEPTANCE
Edit, add or remove criteria as needed. A useful review can include disagreement. An unchanged AI draft is not independent evidence of accuracy.
03 / 07
Influence — DEMATEL
Make your assumptions about influence visible.
How might execution capacity influence financial upside? How much influence runs in the reverse direction?
DEMATEL calculates an influence model from directed judgments. Review each direction, including deliberate zero influence, before calculation. The diagram helps you examine your model of relationships; it does not discover or prove the actual causal structure of the organisation.
Illustrative DEMATEL influence map
Example display threshold > 0.30
Modeled influence sources
Market Size & Growth Rate
Execution Feasibility
Strategic Alignment
Modeled influence
Modeled influence recipients
Financial Upside & ROI
Risk Exposure
Time to Revenue
ASSUMPTION TO CHECK
The example treats execution capacity as an influence on expected returns. Check that relationship against the situation and consider what could invalidate it.
04 / 07
Importance — AHP
Express priorities through comparisons.
How important is feasibility relative to market size for this decision, and why?
AHP turns pairwise preference judgments into weights and an internal consistency check. A consistency ratio of 0.10 or below meets the application threshold. It does not remove bias, resolve stakeholder disagreements or establish that the assumptions are correct. Review any consistency correction before relying on the resulting priorities.
Illustrative weight distribution
Example CR = 0.087
Execution Feasibility & Resource Availability
29.9%
Strategic Alignment & Market Positioning
21.6%
Market Size, Growth Rate & Competitive Landscape
16.5%
Financial Upside & Return on Investment
10.9%
Risk Exposure & Mitigation Complexity
7.4%
INTERPRETING THE MODEL
The example gives feasibility more weight than financial ROI. These weights express the modeled priorities; they are not a universal rule for strategy.
05 / 07
Options & ratings
Inspect the reasoning behind each score.
What evidence supports this rating, and which parts of the explanation are assumptions?
Score each option against each criterion, using higher numbers for better performance. AI can draft ratings and explanations. Check the supporting context, question unsupported inferences, and accept or change the score. An explanation or confidence label is not itself verification.
Illustrative rationale — AIEdge feasibility
Example score: 3 / 10
Illustrative supporting context
"Requires hiring 3–4 senior ML engineers in a highly competitive talent market... engineering already at ~85% capacity... 3–6 month recruitment delays."
Illustrative AI rationale
The initiative requires scarce specialist talent (senior ML engineers) at a moment when the engineering function is already at near-maximum capacity. Hiring delays of 3–6 months on a 26-month runway make this initiative high-risk to initiate in the near term. Rated 3/10 on execution feasibility.
Example rating grid
35 option–criterion scores
Review checkpoint
Inspect, accept or change
HUMAN AUTHORITY
The human retains authority over every judgment. The number of accepted suggestions or overrides describes recorded actions, not how carefully the model was reviewed.
06 / 07
Results & sensitivity
Test how the ranking responds to different assumptions.
Which changes to weights or ratings would put a different option first?
Inspect the deterministic ranking, then explore sensitivity. This illustrative simulation uses 1,000 trials with relative weight perturbations of ±15% before normalization and bounded rating noise with a standard deviation of 0.75 points. Frequencies describe the sampled scenarios. They do not measure the probability that the decision is correct or will succeed in the real world.
Illustrative simulation — 1,000 trials
Weights ±15% / Rating noise SD 0.75
CustomerFortress — Rank #1 Frequency
993 / 1000 trials
99.3%
#1
CustomerFortress
Recommended
6.71
#2
AIEdge (Predictive Analytics)
5.61
#3
PlatformPlay (Partner Ecosystem)
5.29
#4
GeoLaunch (SE Asia Entry)
4.96
#5
DeepExpand (Enterprise Pivot)
4.69
CONDITIONAL RESULT
The example leader stays first in 99.3% of the sampled trials. That frequency is conditional on the selected model and perturbations. Check uncertainties and alternatives outside that scenario set.
07 / 07
Outputs & review
Share the reasoning and the next questions.
A useful report lets others inspect the model, challenge its assumptions and decide what to validate next.
Export the model and decision records, or generate an optional AI-assisted report. Review the version, simulation settings and supporting evidence before sharing. Advisory prose is separate from the deterministic scores. Report contents vary with the saved model and selected report configuration.
Illustrative report outline
Example PDF structure
Executive Summary
Model ranking, assumptions and simulation results
Methodology
AHP weights, DEMATEL maps, CR validation
Alternative Evaluation
Scored comparison across all 5 initiatives
Implementation Roadmap
5 phases, Month 1–24, with go/no-go gates
Contingency Plan
AIEdge activation triggers and hiring sprint protocol
Robustness Appendix
Monte Carlo distribution, sensitivity tables
NEXT HUMAN ACTION
Use the outputs to support a decision discussion. Agree on further validation, ownership and next actions. This example does not establish a typical completion time or savings against consulting.
The model is ready to inspect. Decide what to validate next.
DeciMetrics helps teams make their assumptions, priorities and trade-offs visible. Use the model to support a discussion, inspect disagreements and plan the evidence needed before acting.