Sustainable AI in healthcare: A model from Saudi Arabia

Artificial intelligence (AI) in healthcare has moved beyond proof-of-concept; the question is now how to embed it sustainably as part of a hospital’s digital strategy. Too often, pilots succeed in the lab but still fail in clinical practice – low adoption, poor workflow integration and limited value realization remain the greatest barriers.

Sustainable AI is not about chasing the next algorithm in isolation; it’s about building ecosystems where adoption, governance, and measurable outcomes drive the agenda. This requires a thoughtful and systematic approach focused on delivering real value for patients, providers and health systems.

Saudi Arabian hospital King Faisal Specialist Hospital and Research Centre (KFSHRC) has been developing its own ecosystem over the past several years including a Digital Innovation Hub, covering four key components: Guiding principles, a governance processes, innovation capabilities, and operational discipline.

The following principles and practices are applicable not just in hospitals, but in any organization seeking to implement AI sustainably and at scale.

Rather than starting with technology, successful efforts to implement sustainable AI start with a problem that needs solving, whether that’s a clinical decision-making challenge, a gap in patient care or an operational inefficiency. This “value first, models second” principle keeps initiatives grounded in real needs. In line with this, AI is also embedded into daily routines, not layered on top; ensuring improvement rather than burden for providers and helping to drive adoption.

Tactical principles are also crucial to support implementation. A lean in-house team supported by targeted partnerships can often achieve more than a sprawling AI programme. Infrastructure also matters; a hybrid technical backbone can be ideal, where secure on-premise systems provide the necessary protection for sensitive data, while cloud-based capacity and safe open-source tools allow for scaleability.

Principles should also be considered within the broader ecosystem; for instance when considering innovation partners and commercialization, only a select number of high-value startups should be onboarded within each cycle, with the hospital (or organization) serving as a design partner to ensure their solutions address real healthcare challenges.

While abstract guiding principles are a crucial starting point, it’s crucial to establish repeatable processes that ensure only the right solutions make it into patient care. Therefore, the second component is a clear process for developing new AI solutions and managing a pipeline of innovation is essential. At KFSHRC, governance follows a seven-stage cycle:

Use case initiation: Proposals are accepted year-round but evaluated in structured review windows for clear clinical or operational value, strong feasibility, defined success metrics and evidence from prior work.

Data preparation: Rigorous attention to data quality, lineage and representativeness is applied.

Design phase: Teams decide whether to build in-house, adapt a vendor solution or co-develop with a partner.

Proof-of-concept and proof-of-value: Every solution undergoes a phased evaluation: internal technical testing, external benchmarking and, finally, domain expert validation.

Feedback loops: End-users are engaged to refine usability, reliability, and relevance.

Integration: In addition to embedding into workflows, the process defines follow-up actions. If clinicians or staff do not know how to act on AI outputs, adoption is doomed to fail.

Next steps: Once deployed, solutions enter a structured monitoring phase, tracking utilization, business value and performance. Commercialization pathways including spin-offs and startup collaborations ensure innovation can scale beyond one hospital.

This governance process is designed to be practical: neither a rigid gatekeeping mechanism nor a laissez-faire pipeline, but a balanced pathway that keeps AI safe, impactful and future-ready.

The third key component is to establish a strong capability for generating and scaling AI innovation.

KFSHRC established a Digital Innovation Hub centered on a deliberately lean but highly capable team of fewer than 10 core members, structured to maximize impact while avoiding duplication. The Hub is responsible for product management and data intelligence and has deployed over 30 AI models over the past few years.

Each AI tool is treated as a product from the outset, with roadmaps, success metrics and lifecycle ownership, which ensures solutions evolve with hospital needs. Exploratory data analysis is combined with dataset preparation, enabling AI models to be trained on clean, representative and well-documented data.

Models then enter the AI+ Lab, an innovation engine that develops models, co-designs with partners and pilots AI agents. The Lab experiments to drive new ideas – but within the boundaries of governance. Importantly, innovations are carefully checked at this stage for compliance with national and international standards, ethical frameworks and the internal governance cycle.

Having a multi-disciplinary team helps to deliver and sustain projects, balancing innovation speed with governance depth.

Turning promising pilots into hospital-wide, or indeed organization-wide, impact requires operational discipline: key practices that bridge strategic intent and everyday practice. These levers determine whether AI remains a series of isolated case studies or evolves into a trusted, actionable and sustainable productivity multiplier.

Every AI initiative at KFSHRC requires an internal clinical or operational champion, ensuring collaborations deliver tangible outcomes rather than becoming “innovation theatre.”

This begins with lifecycle accountability. AI solutions are never “delivered and done”; each has a designated owner responsible for updates, monitoring and decommissioning if the system no longer delivers value. This ensures that portfolios evolve over time rather than accumulate unused tools.

Platform thinking further bolsters portfolio management: instead of developing isolated projects, organizations should think about infrastructure – data registries or monitoring dashboards, for example – that makes new models faster, cheaper and safer to deploy.

To support adoption, it’s also crucial to develop applied learning loops. Training is embedded in daily workflows, not just classrooms. That looks like staff participating in iterative feedback cycles, where real-world use informs refinements. In this way, adoption becomes continuous rather than a one-off rollout.

Finally, for larger organizations with the scale to pull it off, curating a select portfolio of startup and vendor partnerships directly aligned with organizational priorities helps operational discipline.

Sustainable applied AI is about alignment: technology, governance, economics, and adoption working together. The future of healthcare belongs to those who build and scale AI ecosystems, not just AI pilots.

Real value in healthcare AI comes from adoption, not experimentation. Leaders must develop strategies that blend internal talent, external partners and clear governance, ensuring innovation efforts generate shared learning and broad impact.

Priorities are sharpening across the system: hospitals should integrate AI into daily workflows; vendors must meet local regulatory needs; and governments should invest in data infrastructure, governance, and workforce capacity to enable scalable progress.

Sustainable applied AI is about alignment: technology, governance, economics and adoption working together. The future of healthcare belongs to those who build and scale AI ecosystems, not just AI pilots.