Why businesses lie about AI

Blog · Read

• Draws on Nikhil Suresh (Hermit Tech) consulting research: in 18 months of enterprise AI consulting, zero successful implementations observed, even as companies publicly claim “100x productivity gains.”
• Frames this as a coordination problem — executives face career risk for admitting AI failure, so individual incentives reward lying regardless of actual results, producing a “religious mania” dynamic where professing faith in AI becomes a promotion/job-security requirement.
• “Demo effect”: executives are unmoved by substantial real technology improvements but get “hypnotized” by unreliable AI demos — cites a Snowflake Cortex demo, described by its own maker as 92% accurate (8% misleading), converting skeptics into immediate buyers.
• Documents “worker theater”: employees completing work manually then crediting Claude/similar tools, or running chatbots in circles solely to rack up tokens for corporate AI-adoption leaderboards.
• Theorizes AI’s appeal to executives specifically: unlike human subordinates, it won’t contradict flawed directives or push back with reality.