Revisit KPIs for AI

KPIs for Articulation

While KPIs are commonly used to monitor performance, they can also be excellent at publicizing value. A data point such as “reduced processing time by 25% this quarter” speaks for itself. It relies on shared understanding, requires no explanation, is inherently shareable as a success story. OKRs and KPIs are the common, and often contractual, language between technologists and stakeholders. So it makes good sense to articulate AI’s contribution through them as well.

What kind of KPIs then? SMART criteria for KPIs are good, but SMART-compliant KPIs can vary widely in narrative power and value signaling.

AI Considerations

Here are three considerations for AI initiatives that have served us well

Economic value KPIs: This avoids the trap of an AI project that seems useful, but cannot help with cost or revenue. We have turned down such projects. Provable economic value of AI is a top stakeholder concern for good reason.

Learning influence on KPIs: Well designed AI solutions include continual learning, which compounds their value over time. Projected KPIs are a great tool to articulate this potential.

Cognitive cycles recovered: AI systems often release human cycles. This is not just more time, but more cognitive focus. This effect is difficult to articulate, but captures a long term advantage. An example is our developers being more focussed on design since there is less load on coding because of coding agents. Under this new distribution of time, they are better designers.

We treat KPIs not just as performance markers, but also as value-signals of where AI is changing the economics and capabilities of the organization. IOW, we think Key Value Indicators.