Part 6/12:
Model evaluation organizations, such as MER (Model Evaluation and Threat Research), track how AI agents perform across tasks. Their findings reveal a remarkable trend: the capability window for AI agents is doubling approximately every 4–7 months—a pace reminiscent of Moore’s Law in semiconductors.
By 2029, AI agents could perform tasks equivalent to one person-month of effort with about 50% success rate, indicating rapid enhancement in complexity handling. This trajectory underscores AI's growing proficiency in managing increasingly sophisticated activities over longer durations.
Practical Applications and Enterprise Impact
AI's transformative potential spans various enterprise functions: