FULL CONFIDENCE. WRONG ANSWER — The Jagged Frontier
AI will win a gold medal at the Math Olympiad.
AI won't reliably tell you what time it is.
That's not a technical anomaly. That's the geometry of the model.
Ethan Mollick named it the Jagged Frontier. The BCG field experiment proved it with 758 consultants: on creative synthesis, analytical writing, and ideation tasks, AI users outperformed by 40% quality, 25% speed. On contextual judgment and real-world reasoning, AI performed worse than no AI. The frontier runs between what the model can pattern-match and what only lived experience can read. Most organizations are operating across that line without knowing where it falls.
✅ The frontier isn't a flaw — it's the architecture.
The same model that synthesizes 3 years of market data in minutes can't read relational irony in a negotiation transcript. It optimizes for statistical pattern, not for judgment built from the real world. Inside the frontier, AI multiplies performance. Outside it, it degrades it — quietly, at scale, with complete confidence. The model doesn't know it crossed the line. That's what makes it dangerous.
✅ The person drawing the line isn't the engineer.
Leading AI adoption doesn't require understanding model architecture. Yet we make frontier decisions daily: which output to trust, which proposal to send, which negotiation to frame with AI. The gap isn't technical knowledge. The frontier is invisible by design — we draw that line without a map. It requires something rarer: the judgment to recognize when the output is technically correct and commercially wrong.
✅ Think globally, act locally — something I learned at Esade, still the most honest commercial doctrine I know — doesn't run on pattern recognition. It runs on the human expertise that reads a specific market, a specific relationship, a specific moment, and decides in real time.
AI doesn't do specific. It does pattern. And in 2026, the medium term has compressed from 1–3 years to 3–6 months. That's not a disruption. That's a different category of problem. The question isn't whether your AI is current. The question is whether you and your organization are genuinely Agile enough to operate at that speed — or whether you've confused deploying AI with building the capability to read what the model will never see.
The frontier doesn't draw itself. Someone in the organization owns that line — and owns what happens on both sides of it.
Unpopular opinion: most AI governance frameworks protect the organization from the model. None protect it from never mapping the frontier. That's not a technology gap. That's a leadership gap — and it compounds at exactly the speed the market is now moving.
In your organization, who decided where AI judgment ends and human judgment begins — and was that a deliberate decision, or did it simply happen?
The Jagged Frontier: where AI ends and human judgment begins
#TheConvergenceTrap — 3.1 of 3
#AIStrategy #CommercialStrategy #OriacGimeno