Testing AI in Corporate Strategy
McKinsey told AT&T in the 1980s that global mobile subscribers would top 900,000 by 2000. The actual number was 100 million – a miss by two orders of magnitude. That forecast was polished, confident, and built entirely on past data: trendslop, long before AI. Today's LLMs produce the same flaw on demand: confident output that sounds insightful until you inspect it. The iron rule from Toyota and AT&T Bell Labs still holds: don't automate what you haven't baselined. Yet most executive teams cannot describe how their own strategy gets made – no baseline exists. The EndPoint Method breaks strategy into six stages; AI is net-positive only in the Snapshot – synthesizing known data. It is destructive in commitment stages: Target Year and Scenario quantification – where LLMs deliver averages, not differentiation. Use AI as a sparring partner, never as decision-maker. Baseline first. Then automate. Your instinct is right: AI can sharpen strategy – and it can wreck it.
Not Rated Yet