The Uncanny Valley Of Enterprise AI Is Bigger Than You Think

Pramata's CEO Praful Saklani reveals enterprise AI outputs look polished but contain critical errors, citing wrong termination clauses, misidentifying counterparties, and contradicting risk flags. The uncanny valley in enterprise AI emerges when outputs appear authoritative yet lack substance, mirroring early robots that seemed human but felt unnatural. Saklani introduces a three-threshold framework to evaluate AI results: first, does it pass the Demo Threshold (theoretical interest)? Nearly all AI tools clear this. Second, does it meet the Accuracy Threshold? Few do. Third, does it satisfy the Substance Threshold? Almost none. These flaws expose a gap between AI's polished presentation and real-world reliability, especially for contract analysis where precision matters.