Overview
- OpenAI’s new evaluation, Measuring the Performance of Our Models on Real‑World Tasks, lists 44 roles where models outperform humans on many tasks, with counter clerks at 81%, editors at 75% and software developers at 70%, while registered nurses show lower exposure at 37%.
- IBM and MIT report widespread worker-led use of AI but weak enterprise integration, with nine in ten employees in Mexico using AI and only 40% seeing effective company use, and just 5% of firms globally moving pilots to production as Shadow AI raises security and compliance risks.
- Marketing continues to post measurable gains from AI, with Statista projecting a $107 billion market by 2028 and McKinsey estimating a 22% lift in digital ad ROI and up to 30% content savings, as WARC notes 72% of brands now operate hybrid human–AI models.
- Practitioners warn that bolting AI onto disordered processes creates costly, unused projects and data‑exposure risks, urging clear problem definitions, staged integration with legacy systems, training and closed architectures for sensitive information.
- Goldman Sachs expects productivity-led GDP growth without proportional job creation, with Gen Z most exposed to entry‑level displacement, while industry leaders caution that outsized investment and expectations could produce notable failures and financial losses.