From simple text generation to autonomously powering robotic arms on production lines, AI technology offers enterprises a kaleidoscope of new capabilities.
So which aspects of AI advancement have companies already embraced, and which ones are on deck for future implementation? The Forbes Research 2025 AI Survey, conducted in August and September, polled more than 1,000 C-suite executives on their AI strategies, including the technologies on their roadmaps.
What’s Being Used Today?
The most widely adopted AI technology is machine learning, deployed by 85% of executives surveyed. However, adoption rates vary by industry:
- The automotive sector leads the way, with 94% of executives reporting that they have implemented machine learning.
- Only 69% in media and entertainment say the same, making it the lowest-ranked industry for adoption.
Generative AI ranks second, with 80% of companies leveraging it. Our global survey uncovered notable regional variations in usage:
- Asia-Pacific: 92%
- Europe, the Middle East and Africa: 89%
- Latin America: 80%
- North America: 70%
Those two are followed by image recognition, natural language processing (which allows computers to understand written and spoken human language) and predictive AI, which can make forecasts.
At the other end of the spectrum, robotics is the least deployed AI technology overall, used by just 21% of respondents overall. But in specific industries, usage rises significantly:
- 51% in automotive
- 46% in manufacturing
What’s On The Horizon?
When asked which AI technologies they plan to deploy in the future, executives listed multi-agent systems as their top choice (62%). These systems involve multiple AI agents working together to solve complex problems.
That was followed by reinforcement learning (60%), a machine learning subfield focused on training an AI agent to make optimal decisions based on feedback. At third spot was federated learning (53%), which trains an AI model on decentralized data without moving the data to a central server. This privacy-forward approach is useful, for example, when training healthcare AI models to detect tumors without compromising patient privacy.
