Accountability
AI investment decisions should assign responsibility for expected benefits, operational failures and ongoing governance costs. Value claims should be checked against actual outcomes rather than ending at deployment approval.
Business & innovation IWIB4AI session Needs review
Power Morning Coffee with Alexander Shepherd – Rethinking AI ROI
Measuring AI investment value through governance, regulatory readiness and technical quality, rather than financial returns alone.
Draft – legal and case law references not yet verified
Summary
This use case examines how organisations evaluate AI investments when conventional return-on-investment metrics overlook governance, regulatory obligations and technical foundations. A broader assessment considers financial performance alongside reliability, compliance costs and impacts on affected people. It treats governance by design as part of delivering lasting value, rather than merely an additional expense.
The Problem
Short-term financial metrics can reward rapid deployment while obscuring unreliable performance, downstream harms and ongoing compliance costs. The challenge is to measure AI value without treating legal obligations or fundamental rights as trade-offs that financial gains can outweigh.
Ethical Grounds
Transparency
AI business cases should disclose the assumptions, uncertainties and limitations behind projected returns. Decision-makers need visibility into technical dependencies and compliance costs to compare investments meaningfully.
Safety
Technical reliability and risk controls should form part of investment assessment. Financial savings alone cannot justify deploying a system whose foreseeable harms remain inadequately controlled.
Fairness
Value measurement should examine who receives the benefits and who bears errors, costs or additional burdens. Aggregate returns can conceal unequal impacts on workers, customers or other affected groups.
Legal Compliance
| Framework | Provision | Relevance |
|---|---|---|
| EU AI Act | Arts. 9 and 17 | For providers of high-risk AI systems subject to these requirements, risk management and quality management create lifecycle obligations that should be built into investment planning and resource allocation. |
| GDPR | Art. 25 | Where AI processes personal data, controllers must implement data protection by design and by default, making appropriate safeguards part of the technical and financial plan. |
| GDPR | Art. 35 | AI processing likely to result in a high risk to individuals’ rights and freedoms requires a data protection impact assessment before processing, which should inform project feasibility and safeguards. |
| EU AI Act | Art. 4 | Providers and deployers must take measures to ensure sufficient AI literacy among relevant personnel, making competence development a necessary consideration in AI investment planning. |
Sources
Court & Regulator Decisions
NJCM et al. v State of the Netherlands (SyRI) (opens in a new tab)
The invalidation of welfare-fraud profiling illustrates why investment assessments must examine rights compatibility and safeguards, not merely projected efficiency gains.
Moffatt v Air Canada (opens in a new tab)
Liability for chatbot errors demonstrates why AI investment assessments should include reliability testing, remediation costs and responsibility for downstream customer harm.
Key Takeaways
- Use a value scorecard covering financial returns, technical reliability, governance and impacts on people.
- Budget for lifecycle costs, including data quality, monitoring, training and corrective action.
- Map applicable legal obligations early and treat compliance as a baseline, not an optional financial trade-off.
- Assign owners to value claims and reassess outcomes against a clear baseline after deployment.
Related Use Cases
From the Session
Power Morning Coffee with Alexander Shepherd – Rethinking AI ROI
Speakers
