AI in 2025: current initiatives and challenges in large enterprises

Traditional AI technologies like Machine Learning and Computer Vision have been implemented for decades, while GenAI marks a new phase, enhancing productivity and personalization. Companies are moving from proofs of concept to large-scale deployments, focusing on aligning business and technology goals. Demonstrating ROI remains complex, requiring clear metrics. Change management is crucial for adoption, addressing barriers like tool usability and perceived reliability. Balancing innovation with strategic planning ensures technological advancements align with organizational goals.

Enterprises must balance traditional AI and GenAI solutions, tailoring mixes based on specific use cases. Centralized platforms connect company data, while separate pipelines address business-specific needs. MLOps and LLMOps streamline development of AI systems, for enhanced scalability and reliability. Companies face the «Make or Buy» decision for AI models, balancing cost and customization. The future of AI is multi-modal and multi-agent, integrating voice, image, video, and text interactions to tackle complex challenges, drive innovation, and enhance operational efficiency across various sectors, ensuring seamless integration and optimal performance.

AI intensifies existing risks and introduces new threats, requiring proactive measures. Companies must combine traditional security with innovative defenses to safeguard assets. Emerging frameworks, including the EU AI Act and the NIST AI RMF, guide organizations on cybersecurity and privacy. Ensuring an adequate security posture and managing incidents involving AI models are increasingly challenging and oftentimes outsourced for the time being. Adopting a risk-based management approach balances innovation with risk awareness, enabling companies to navigate the evolving landscape of trustworthy AI effectively.

AI democratization raises ethical issues, especially with GenAI, impacting jobs and requiring ethical guidelines. Companies must address inclusion, bias, privacy, intellectual property among other topics. GenAI’s environmental impact is significant due to energy and water consumption. Companies try to promote frugal practices and green engineering strategies, like Small Language Models and environmental impact measurement. There is, nevertheless, a lack both of transparency from providers and of metrics to better measure impact. Balancing technological advancements with ethical and environmental considerations is essential for sustainable AI practices.

Traditional AI has already revolutionized business processes in the past decade, and GenAI is about to do the same with a broader workforce. Today, levels of adoption vary, with challenges in bridging knowledge gaps. Companies are deploying «Secure GPT» systems and scaling AI solutions across organizations, but the tech environment keeps changing at a rapid pace. HR must anticipate changes, redefine objectives, and address workforce impacts. Advanced training and secure AI solutions are crucial for transformation. GenAI is a catalyst for reimagining work, fostering innovation, and attracting talent, ensuring employees harness these technologies responsibly and effectively.

AI drives a global competition for talent, with significant salary and availability differences across regions. Traditional sectors and startups compete with tech giants for AI professionals. The talent shortage is due to the rapid pace of AI innovation outstripping education and training programs. Upskilling and reskilling strategies are essential. Retention involves cutting-edge projects, top technologies, and links with academia. Enhancing AI literacy among decision-makers is also vital. Collaboration between academia and industry, along with governmental support, can foster a robust AI ecosystem.

The 21st century’s global competition in AI is led by the US and China, with tech giants extending national influence. Europe lags, risking technological dependence. Companies must pursue strategic autonomy, aligning public and private interests. A clear industrial AI policy is needed, fostering local ecosystems across key sectors. Open-source AI models are also a vector of innovation and inclusiveness, by making the latest tech available to everyone. A global AI governance might be difficult to enforce in a polarized world, but seems essential to balance national and private interests, ensuring technological advancements benefit the majority.