Page 44 - AI for Good Innovate for Impact

AI for Good Innovate for Impact

3. Learnings and conclusion

3�1 Cross-Sector Insights from Global Use Cases

This report presents a synthesis of 160 real-world AI use cases across eleven critical domains,
offering a comprehensive view of how AI is being applied to address societal challenges,
drive innovation, and enhance efficiency. It highlights both the transformative impact of AI
across sectors and the persistent gaps—technical, governance and infrastructural—that must be
bridged to achieve scalable, inclusive, and sustainable deployment worldwide.

1. An analysis of 27 healthcare AI use cases highlights three key barriers to large-scale
adoption: lack of data standardization, limited real-world deployment of edge AI, and
insufficient explainability in multimodal models. Despite these challenges, the field is
advancing rapidly with trends such as specialized large models, generative AI for medical
data, federated learning for privacy-preserving training, and lightweight AI for mobile
health. Some models now match or surpass expert performance, especially in multi-
cancer detection, signalling a shift toward more intelligent, inclusive, and deployable
AI-driven healthcare solutions.
2. Over 23 AI use cases in climate and disaster response reveal major challenges, including
inconsistent data standards, regional data scarcity, and limited global scalability.
Nonetheless, the project has helped collect valuable data from underrepresented
regions, enhancing model generalization and supporting more locally tailored, effective
solutions. Through cross-border collaboration and data sharing, AI is driving a more
inclusive and resilient global response to climate risks.
3. AI use cases in the 5G domain face three key challenges: infrastructure integration across
heterogeneous systems, real-time performance limitations, and compliance and security
risks in cross-domain applications. Still, a dual transformation is underway. "Network
for AI" leverages 5G/6G infrastructure—cloud-edge collaboration, network slicing, and
programmable interfaces—to support scalable AI workloads. In parallel, "AI for Network"
enables intelligent operations like predictive maintenance and dynamic resource
management, reshaping telecom infrastructure into an adaptive, self-optimizing platform.
4. More than 20 AI productivity use cases illustrate a shift from experimental development
to real-world deployment. These applications—from smart HR systems to multilingual
translation tools—feature mature architectures, scalable integration, and growing
adoption. While challenges around data quality and localization persist, they are being
addressed through human-in-the-loop methods and domain adaptation. Trends like API
integration and open-source adoption signal that AI is now a key driver of cross-sector
productivity.
5. Current AI use cases in manufacturing highlight the convergence of AI, IoT, and robotics to
enable smart inspection, autonomous operations, and predictive maintenance. However,
challenges remain in communication reliability, data availability, safety standards, and
trust in automation. Addressing these issues is essential to realize intelligent, adaptive
industrial ecosystems at scale.
6. Financial AI use cases raise concerns about the balance between personalization and
privacy. Intelligent services such as robo-advisors and voice-activated banking depend
on sensitive user data yet often lack transparency and adequate control mechanisms. Best
practices—including on-device AI, differential privacy, federated learning, and explainable
AI—are emerging to address these gaps and build secure, ethical, and user-centered
financial services.
7. Many AI education solutions remain in pilot phases, constrained by scalability issues,
infrastructure gaps, and misalignment with local curricula. While tools like AI tutors and

8