AI GUIDE FOR BOARDS IN SINGAPORE
Section II: Strategy and Value Creation
3. Model performance: Accuracy, reliability, robustness, precision and recall, dri , hallucination rates, system availability. 4. Adoption and usage: User uptake, frequency of usage, repeat engagement, workow integration, productivity impact. Leading indicators may include employee adoption, workow experimentation and training completion. Lagging indicators may highlight revenue outcomes, e ciency gains and margin improvement.
In early stages, pilot initiatives are typically used to validate technical feasibility, user acceptance and practical usefulness. At this stage, metrics should focus on technical performance and initial adoption; nancial outcomes may be limited or not yet measurable. As initiatives scale operationally, boards should expect increased emphasis on business integration, adoption, productivity gains, nancial and operational impact. At the enterprise level, metrics must also reect business outcomes, strategic value creation and operational resilience. A staged approach ensures that early experimentation is not prematurely assessed on nancial outcomes, while scaled initiatives are held accountable for delivering measurable business impact. Metrics that are appropriate for experimentation may be insu cient once AI systems are scaled across the enterprise. Boards should ensure that management applies appropriate metrics at each stage of maturity.
5.1.3 Anchoring metrics to business context
Boards can make use of performance indicators (See: AI Performance Indicators: Examples ) to reect the speci c operational context in which value is created. Management should de ne key performance indicators that are directly linked to the underlying workow, business problem, or customer outcome. Industry-speci c examples include: • Customer operations: Metrics should focus on customer resolution time, rst- contact resolution rates, call containment rates, escalation frequency and customer satisfaction. ese indicators help assess whether AI is improving operational e ciency and customer experience. • So ware development: Indicators such as developer productivity, code quality, testing e ciency, defect reduction, deployment frequency and speed of iteration can help capture gains in delivery capability and service reliability. • Knowledge and decision-support: Monitoring retrieval quality, response relevance, search success rates, reduction in duplicated work and decision turnaround time can help assess whether AI is improving information access and decision-making eectiveness.
5.1.2 Using a balanced measurement framework
Boards should encourage management to adopt a balanced measurement framework incorporating multiple dimensions of AI performance. e review can start with the adoption of a balanced set of indicators across four broad categories: 1. Financial value: Return on investment, cost savings, revenue upli , margin improvement, productivity gains. 2. Operational impact: E ciency gains, cycle time reduction, automation rates, throughput improvements, faster response time, reduction in operational bottlenecks.
44
Made with FlippingBook Digital Proposal Creator