AI in the Hospital Needs a Policy Before a Pilot
AI's potential in hospitals is no longer in doubt. The hard part is putting it into daily practice safely, and that starts with governance, not with a pilot.
I began using generative AI in my own work as a hospital chief executive within weeks of ChatGPT's release at the end of 2022. It quickly became useful for reviewing data, drafting communication, planning and following up. Since then, AI has moved from curiosity to boardroom agenda in almost every health system in the Gulf. What has not moved as fast is the ability of hospitals to adopt it well.
Why this matters
The obstacle is not the technology. It is implementation. Hospitals are buying or piloting AI tools without clear rules on who approves them, what data they may use, how their outputs are checked and who is accountable when they are wrong. That creates two risks at once: unsafe use in some places, and paralysis in others where staff do not know what is allowed.
In July 2024 I hosted a LinkedIn audio event, "AI in Healthcare: Let's Talk Transformation", with Adel Al-Shabaan of Health.Links and Khalid Almaghrabi of Fakeeh Technologies. We discussed AI already in use at Dr Soliman Fakeeh Hospital in Jeddah, including predictive models to improve appointment scheduling and AI-supported personalisation in patient communication. We also discussed data bias, the explainability of models, cybersecurity and organisational readiness. The conversation ended where most serious conversations about AI end: with governance.
Where to start
The most mature regulator-cleared AI tools today are in diagnostics, especially imaging. Clinical use rightly requires strict validation and oversight. The fastest, lowest-risk returns are usually found first in non-clinical areas: scheduling, revenue cycle, documentation support, operational dashboards and patient communication. These build confidence and capability before AI touches clinical decisions.
A five-step plan
1. Write an AI policy. Define what AI may and may not be used for, what data it can access, how outputs are checked and how patients are informed.
2. Form an AI taskforce. A small, senior, multidisciplinary group, including clinicians, IT, legal and quality, that reviews proposals and owns the portfolio. Whether that eventually needs a Chief AI Officer is a fair question for larger systems.
3. Build adoption. Train staff at every level. AI will not replace clinicians, but clinicians who use it well will increasingly outperform those who do not.
4. Choose first use cases deliberately. Pick areas where benefits can be measured and risks are low, and prove value there first.
5. Measure and report. Track impact and safety, report to the board, and stop what does not work.
Questions for boards and leadership teams
- Do we have a written AI policy, and does every member of staff know it exists?
- Who approves a new AI tool before it is used with patient data?
- Which AI initiatives have measurable results, and which are still demonstrations?
AI should give clinicians time back for judgment and for patients. It should never replace either.
Further reading: The Coming Wave of AI in Healthcare · Connected Care (connected.care).
Dr Fatih Mehmet Gul is a physician and healthcare executive who has built, commissioned and led hospitals across the Gulf and Türkiye. He is an Honorary Professor at UCL Global Business School for Health, author of Connected Care, and host of The Chief Healthcare Officer podcast.
More insights →