3.5 · The ROI Engine — What's AI Worth to Your Business?Download
Course 3 Key Takeaways
⏱ 5 minCourse 03
A summary of the core frameworks and decision tools from Course 3 — Strategic AI Decision-Making.
Your Course 3 Reference
The Five-Dimension Framework
- ◆Score 1–4 on: Strategic Fit, Data Readiness, Operational Impact, Risk Exposure, ROI Viability
- ◆Score 17–20: Proceed. 12–16: Conditional. 8–11: Redesign. Below 8: Decline.
- ◆Any single dimension scoring 1 triggers mandatory board escalation
The Autonomy Spectrum
- ◆Level 1–2: Recommended starting point for all new AI deployments
- ◆Level 3: Co-pilot — AI handles routine, escalates complex cases
- ◆Level 4+: Requires sustained performance track record and formal governance
- ◆Autonomy is easy to grant and hard to retract — design oversight first
Red Lines
- ◆Legal red lines: EU AI Act prohibited practices (in force Feb 2025)
- ◆Emotion recognition in workplaces and educational settings is prohibited
- ◆GDPR Article 22: Automated decisions with significant effects require human review capability
- ◆Organisational red lines: document what you won't do beyond what's illegal
The Eight Gates
- ◆Gate 3 (Ethical Review) is where most proposals should be challenged — but rarely are
- ◆Gate 8 (Post-Deployment Monitoring) is the most commonly skipped — and the one that catches slow failures
- ◆A gate failure means redesign or decline — not bypass
- ◆Well-prepared proposals should pass in 3–6 weeks
ROI Calculation Principles
- ◆Start with a documented baseline — no baseline, no credible ROI
- ◆Apply a 50–60% realisation factor to vendor-quoted automation rates
- ◆Include full implementation costs: vendor, data prep, integration, training, maintenance
- ◆Model the ramp-up curve — full savings typically arrive at month 7+
