AI hits low bars, fails quality tests

AI is currently a "disenchanted intern" capable of routine drafting but failing at high-stakes, multi-step professional work.

• –The "Iceberg Index" reveals AI technical capabilities extend to 11.7% of the labor market, yet visible adoption remains at just 2.2% due to quality gaps
• –MIT's simulation of 151M digital twins reveals a "complexity gap" where AI rarely achieves superior, error-free output for tasks requiring multiple steps
• –Deloitte's fabrication of citations in Australian and Canadian government reports serves as a critical warning for firms prioritizing cost-cutting over accuracy
• –Performance is improving at 11% annually, suggesting "minimal sufficiency" for most tasks by 2029, yet "superior" quality remains the human moat