AI pilots do not become operating value
Teams experiment with assistants or agents in isolated functions without redesigning the underlying work. This becomes most visible when pilots move from assisting individuals to making or executing decisions inside operational workflows.
The technology sits beside the operating model rather than changing it. The underlying weakness is the absence of an operating model that defines purpose, authority, constraints, evidence, escalation and revocation for human, agentic and hybrid actors.
Pilots multiply, governance fragments and EBIT/productivity impact remains elusive. At enterprise scale, experiments remain isolated or scale without consistent accountability, producing unclear value and unacceptable governance exposure.
Map capabilities, redesign the work and progress autonomy deliberately rather than deploying agents opportunistically. In practical terms, each capability can progress to the autonomy level justified by its value, risk and human-oversight requirements under Human Sovereignty.
- Fewer disconnected pilots; stronger workflow-level value; clearer path from experimentation to enterprise scale
- Clear autonomy and authority boundaries
- More credible AI value cases
- Continuous evidence and oversight
- Controlled progression from assistance to autonomy
- Which AI pilots changed an end-to-end workflow?
- What operating-model constraint prevents scale?
- How are pilot outcomes compared across capabilities?
- What evidence must an agent produce before its action is accepted?
- Under what condition must authority return immediately to a human decision-maker?
McKinsey reports widespread experimentation with AI agents but much narrower scaling, usually limited to one or two functions.
McKinsey: The State of AI 2025
NIST provides a use-case-and risk-based framework for governing, mapping, measuring and managing AI risk across the system lifecycle.
NIST: AI Risk Management Framework