The Five-Dimension Framework
The Five-Dimension Framework evaluates any proposed AI deployment across five dimensions. Each dimension is scored 1–4. A total score guides the deployment decision — but the framework also flags individual dimension failures that can veto a deployment regardless of the overall score.
Dimension 1: Strategic Fit
Does this AI use case directly serve a documented business objective? The test: Can you name the specific KPI this AI will move, and by how much? If the answer is vague — "it will improve efficiency" — the strategic fit is weak. Score 4 if the AI maps to a specific, measurable objective. Score 1 if the use case exists because "everyone is doing AI".
Dimension 2: Data Readiness
Does your organisation have the data quality, volume, and governance to train and run this AI reliably? This is the dimension most organisations underestimate. The test: Is your training data representative of the full range of situations the AI will encounter? Is it labelled, clean, and legally obtained? Can you retrain the model as the world changes? Score 4 if data infrastructure is in place. Score 1 if you're planning to "sort the data out once the model is built".
In most failed AI projects, the data problem was visible before the project started. The organisation chose to proceed anyway, assuming the data would "be good enough". It rarely is. Data readiness must be assessed honestly, before a single pound is spent on model development.
Dimension 3: Operational Impact
The test: What is the worst-case outcome if this AI makes a mistake? Who is affected, how severely, and how often might it happen? An AI that recommends slightly suboptimal marketing copy has low operational impact. An AI that flags medical risk scores has high operational impact. The higher the impact, the more robust your oversight mechanisms need to be.
Dimension 4: Risk Exposure
What regulatory, reputational, or legal risks does this deployment carry? The test: Have you checked this against the EU AI Act risk tiers (from Course 2)? Have you assessed GDPR implications for personal data? Have you considered what would happen if this system's outputs appeared on the front page of a newspaper? Score 4 if you've done a formal risk assessment. Score 1 if risk hasn't been formally considered.
Dimension 5: ROI Viability
Is the investment proportionate to the problem, and can the return be measured? The test: Can you calculate — not estimate, calculate — the expected return within 12 months? Is there a clear baseline measurement? Is the cost of implementation justified by the problem it solves? We'll go much deeper on this in Section 5's ROI Engine. For now, the key principle is: if you can't measure the value, you won't be able to justify the project — or know when it's working.
Using the Framework: Scoring and Decisions
- ◆Score 17–20: Strong deployment candidate. Proceed with pilot design.
- ◆Score 12–16: Conditional. Address the weakest dimension before proceeding.
- ◆Score 8–11: Redesign required. The use case may be worth pursuing with significant changes.
- ◆Score below 8: Decline or defer. The conditions for responsible deployment are not met.
- ◆Any single dimension scoring 1: Automatic escalation — do not proceed without board-level review.
The Five-Dimension Framework is a structured tool, not a bureaucratic checklist. Its purpose is to surface the questions that deployment pressure tends to suppress. Use it as a conversation guide with stakeholders — the discussions it generates are as valuable as the score it produces.
