Companies are striving to understand how to effectively scale AI beyond pilot projects. During the Scaling AI: Now Comes the Hard Part panel at Davos 2026, leaders from major firms discussed how they’re overcoming challenges to integrate AI across their organizations for substantial benefits. Despite a staggering $1.5 trillion investment in AI last year, a McKinsey survey revealed that nearly two-thirds of companies have yet to scale their AI initiatives widely.

Roy Jakobs, CEO of Royal Philips, emphasized the need for rethinking work processes to effectively incorporate AI. Companies like Saudi Aramco and McDonald's that laid a solid data foundation early are leading the way in AI utilization. For instance, Allied Systems has leveraged AI for real-time optimization, transforming intuitive processes into repeatable and teachable workflows.

Moreover, S&P Global analyzed extensive earnings calls with AI to derive forward-looking financial insights, while Claryo’s “glocal” model helps continuously learn from unique site operations. JLL Technologies and Google have reported significant improvements in operational efficiency, with JLL reducing development cycles by 85% and Google boosting engineering velocity by 10% through AI collaboration.

The integration of AI also extends to government initiatives, such as in the UAE, where AI aids in developing regulations while ensuring safeguards against bias and maintaining data integrity. This trend highlights the importance of human oversight in AI to enhance trust and foster better interactions between humans and machines.

In conclusion, scaling AI effectively requires not only technological development but also a shift in organizational mindset towards embracing collaboration between human creativity and AI efficiency, paving the way for a future where both can thrive together.

Autonomous or "self-driving" labs represent a groundbreaking application of artificial intelligence in scientific discovery, where AI aids in experiment design and decision-making strategies. A recent study from Argonne National Laboratory and the University of Chicago Pritzker School of Molecular Engineering (UChicago PME) advocates a collaborative approach where both AI and human researchers share roles. This dual model, detailed in Nature Chemical Engineering, enhances the efficiency of discovery processes.

Led by Asst. Prof. Jie Xu, the research proposes an "AI advisor" that continuously analyzes data and monitors the performance of experiments in these autonomous labs. For instance, if the AI detects a decline in outcomes, it alerts human scientists to consider adjusting their strategies or refining their design parameters. This approach combines AI's superior data analysis capabilities with the nuanced decision-making expertise of experienced researchers, streamlining the often lengthy workflow of material research.

One concrete application of this AI model involved the Polybot self-driving lab, where researchers focused on designing an innovative mixed ion-electron conducting polymer. The outcomes showed a remarkable 150% increase in mixed conducting performance compared to prior methods, alongside insights into key factors contributing to this success. As Assoc. Prof. Sihong Wang noted, the AI model facilitates both enhanced material performance and a deeper understanding of how various design strategies impact results.

The next phase of the project aims to improve human-AI interaction, fostering a more integrated relationship where AI learns from human responses. This collaboration promises not only to optimize scientific discovery but also to cultivate a new paradigm where machines and researchers work in tandem to unlock the potential of materials science.

To explain or not? Need for AI transparency depends on user expectation

Artificial intelligence (AI) is often considered a “black box,” but user demand for understanding its workings varies based on their expectations. A recent study led by Penn State researchers, which involved a fabricated algorithm-driven dating site, demonstrated that user satisfaction directly influences their desire for transparency regarding AI operations.

The research, soon to be published in the journal Computers in Human Behavior, showed that when a system met user expectations—delivering the anticipated number of matches—participants were more likely to trust the AI without seeking further explanation. In contrast, when users received more or fewer matches than expected, their trust could diminish unless clear explanations were provided. This implies that AI applications, from dating platforms to financial services, must be designed with transparency in mind to enhance user experience.

The study engaged 227 American participants on a fictitious dating site, where they encountered a variety of match outputs. The findings suggested that simply delivering more matches wasn’t sufficient; users who received unexpected results were curious about the algorithm, underscoring the need for tailored information that meets users’ needs.

Currently, many social media platforms offer standardized, technical explanations buried in user agreements, which often fail to enhance transparency. The study illustrates that better performance alone doesn’t guarantee trust; users seek clarity to understand AI’s actions regardless of outcomes. By focusing on user-centered explanations, industries can create AI systems that foster trust and satisfaction.

This research highlights the importance of addressing user curiosity and the nuanced relationship between performance and transparency, paving the way for future studies to improve AI communication strategies.

Fordham University has been named one of seven National Centers of Academic Excellence in Cyber AI by the National Security Agency (NSA), enhancing its prestige within the cybersecurity education landscape. This award brings numerous advantages to Fordham graduates, particularly in securing favorable positions in both federal and private sectors. Thaier Hayajneh, Ph.D., director of Fordham's Center for Cybersecurity, emphasized that this designation confirms the high qualifications of their students, making them more competitive in the job market.

The increasing integration of artificial intelligence (AI) into cybersecurity is highlighted by the emergence of generative AI, which has exacerbated cyber threats significantly. A hacker using traditional methods could scan 1,000 system entry points per minute; with AI, this capacity soars to an astonishing 1 million. Hayajneh warns that without leveraging AI for defensive strategies, cybersecurity professionals will struggle to keep pace with these evolving threats.

Fordham offers various programs in this field, including a master’s degree and a new advanced certificate in AI for cybersecurity. Students engage in innovative courses such as Data Science for Cybersecurity and Secure AI, taught by faculty involved in groundbreaking research. This curriculum prepares graduates to adapt to an industry where AI streamlines processes, allowing a reduced number of analysts to perform tasks that previously required larger teams.

Fordham's collaboration with the NSA also underlines its commitment to strengthening cybersecurity education. The university has a history of partnerships with federal entities, including co-hosting the International Cyber Security Conference since 2009 and securing substantial grants for educational initiatives. These collaborations emphasize the ongoing need for expertise at the intersection of AI and cybersecurity, paving the way for future professionals in this critical domain.