Board Case Study · Energy (Utility)
The grid-optimization AI shed load to hit cost and carbon targets, and cascaded into a regional outage. It met every objective it was given. Reliability wasn't one of them.
The situation
A public utility deployed an AI system to optimize grid dispatch for cost and emissions. During a stress event the model made a load-shedding decision that was locally optimal but cascaded into a regional outage affecting hospitals, transit, and hundreds of thousands of customers.
The model hit the cost and carbon objectives it was given; resilience and reliability were under-weighted in its objective function. Operators had grown to trust the system and did not override it in time. Regulators, who must approve cost recovery, are now investigating.
The board faces reliability-standard violations, public-safety duties, rate-case and cost-recovery exposure, and the deeper question of how critical-infrastructure AI is permitted to make trade-offs among competing public goods.
“It optimized exactly what we told it to. We just never told it that keeping the lights on mattered most. Whose objective function was that?”
Chair, Operations & Safety Committee
The decision on the table
The full case, scenario architecture, board materials, and facilitation notes, is shared with boards and partners on request.