Source copy companion for the smooth-system visual briefing.
The most dangerous AI systems aren't the ones crashing. They are the ones operating perfectly.
For twelve years, I managed data and reporting for complex human systems, including the resettlement infrastructure for Mozambique LNG. If there is one universal truth about managing operational risk, it is this:
Human reality is high-friction. If the dashboard is smooth, the system is lying to you.
Right now, operational leaders are deploying AI to manage supply chain compliance, worker grievances, and social risk intake. The vendors promise a "frictionless" experience. The AI ingests messy, dialect-heavy, emotionally charged human input, and spits out clean, categorized summaries. The charts go green. The board is happy.
But you cannot automate away human friction; you can only displace it.
When an AI system forces a complex human reality into a clean drop-down menu, it creates a "Smooth System." To achieve that smoothness, the AI quietly flattens the nuance. It ignores the non-standard dialect. It hallucinates a polite summary of an unprecedented threat.
The friction doesn't disappear. It builds up off-ledger. It becomes unquantified decision risk. And eventually, it detonates—usually in the form of a labor dispute, a compliance failure, or a PR crisis that "nobody saw coming" because the dashboard said everything was fine.
If your AI architecture prioritizes a frictionless dashboard over the preservation of messy, high-friction human reality, you are not managing risk. You are hiding it.
(If you are an operations or ESG leader accountable for what happens off-ledger, I run confidential, fixed-scope Systems Briefings to find the friction your AI is currently hiding. Link in the comments.)
