Health systems look to scale AI beyond the pilot phase

By Bill Fera, M.D., principal, and Haleigh Sinkewich, senior manager, Deloitte Consulting LLP

The use of artificial intelligence (AI) in health care is still in its early stages. But some health systems have moved beyond pilot projects and are scaling use cases. Some early adopters have demonstrated improvements, such as reduced diagnostic errors, optimized treatment paths, and streamlined operations.¹

We recently spoke with Luis Taveras, Ph.D., senior vice president and chief information officer (CIO) at Jefferson Health in Philadelphia. For the past year, he has been navigating the organization’s implementation of AI and is beginning to expand from the pilot phase into broader implementation. We recently had an opportunity to talk with Luis about the potential of AI in health systems, the importance of governance, and how to move beyond the implementation phase to scale AI throughout the organization. Here is an excerpt from our conversation:

Deloitte: When introducing a new technology like AI, it can be easy to get stuck in the prototyping stage. What role do you think governance can play in identifying use cases and effectively scaling them?

Luis: There is a governing process for moving an idea into a project. People sometimes think that any good idea they have should become a project. I always want our people to look for new ideas, but those ideas need to go through a process before they can be considered as a new project. It’s important to manage the whole process from ideation…very few ideas actually become a project. Most health systems already have a governance model. They may just need to figure out how AI is going to fit. You usually don't need to create a whole new governance for AI. However, some health systems don’t have a governance model at all. In that case, the adoption of AI could be the catalyst for creating one.

Deloitte: Health systems typically receive a lot of information (and proposals) from IT vendors. How should they manage that?

Luis: It can be challenging. Electronic Medical Records (EMR) vendors usually have salespeople who reach out to key people within a health system. We have to be sure that our people understand there is a process for evaluating technology like AI. In addition, there are more rules that need to be followed in health care compared to other industries. We have to protect patient information, which means we can't provide a vendor with patient data to train their large language models. We are dealing with life and death, so we can’t just experiment with new technologies.

Deloitte: There are expectations that AI will help reduce the amount of time clinical staff spend on paperwork so that they can spend more time providing direct care to patients. Are you seeing that?

Luis: Yes…in our ambient speech pilots. Our physicians are now spending less time in front of a screen, and more time interacting with patients. We are tracking the productivity of clinicians and measuring the amount of time they spend with patients. The other true measurement is how much free time the physicians have once they go home. Not having to finish charting at home means they have more time to spend with their families, and that can help reduce burnout. We use AI for ambient documentation and are moving beyond the pilot phase into a rollout. We are working with medical staff leadership to determine who will have access to the technology because we have a limited number of licenses, and we need to maximize the value of those. There are some areas in clinical care where there is little interaction between the patient and the physician, so ambient documentation doesn't make much sense. And some physicians need to dictate their post-surgery notes. But in a primary care office, ambient listening could be a great help. We look to the medical staff leadership to guide us. The use of AI in documentation could result in less burnout and better retention among clinical staff, which can help reduce expenses associated with turnover among staff.

Deloitte: Can AI help improve efficiencies among clinical staff?

Luis: Clinical staffing is a key issue for many health care organizations. Another issue is rising costs and shrinking budgets. I think AI has a role to play in driving efficiencies across health care organizations without increasing costs. For example, there is a shortage of radiologists in our system and across the industry; and radiologists usually have to read a tremendous number of images.² AI, for example, can be used to do an initial read of mammography images to help identify those that might have an issue and need to be prioritized. This can help focus the radiologist’s review. The result can be a quicker diagnosis for patients, which is critical when you're dealing with cancer. We started the implementation by assembling the radiology leaders and asking them to identify areas where there is duplication or overlap. Now we are trying to apply some of what we’ve learned in radiology to other areas, like pathology. Could AI help make pathologists more efficient by limiting the number of specimens they need to evaluate?

Deloitte: Is there a danger that AI could replace some clinical staff?

Luis: In health care, it is important to view AI as augmented intelligence, rather than artificial intelligence. This means, everything starts with a clinician and ends with a clinician. If the AI model determines an image is normal, that determination still has to be validated by a radiologist. AI can make the process more efficient, but clinicians still need to be involved in the front and the back end of the process.

Deloitte: Have you seen collaboration with CIOs across health systems while implementing AI?

Luis: The collaboration with regional CIOs and national CIO organizations has been significant. There seems to be an acknowledgement that AI could benefit the well-being of our medical staff. At present, there is no return on investment (ROI) in terms of hard dollars, but I think this technology is going to be transformational when it comes to the medical staff. Governance is important to drive the deployment of these solutions and make sure we have a way to measure the value. We should measure the value, because ROI will likely drive this going forward.

Deloitte: Why do you believe hospitals should consider adopting AI?

Luis: I believe they may benefit from efficiencies and improved patient care that can come with AI. For example, we are evaluating AI solutions that can predict when a patient is at a high risk for falling out of bed. This solution may be able to give the clinical staff a 45-to-60-second warning that a patient is likely to fall…and that gives staff time to try to prevent it from happening.

Deloitte: How do you think AI could be used in health systems five years from now?

Luis: There may be many questions and implications ahead for AI. You need to consider how AI might transform health care. Will patients use AI to accurately self-diagnose themselves? If that happens, what happens to primary care?³ What happens to the diagnostics that are used in primary care? Will a patient be able to self-diagnose and go right to a specialist for treatment? How are outcomes impacted when a nurse doesn’t have to wake a patient up multiple times a night to check vitals? AI solutions are still expensive, although not as expensive as they were 18 months ago. I think that the cost curve is going to continue to come down, and AI solutions will become more affordable for rural and smaller organizations.

Conclusion
The integration of AI into health care appears to be moving from theoretical potential to real-world application. Through thoughtful governance, strategic collaboration, and a focus on improving both clinician efficiency and patient outcomes, AI can deliver meaningful value—particularly in areas like documentation, diagnostics, and operational efficiency. However, the path to widespread adoption is likely not without challenges. Issues such as data privacy and cost can be hurdles. Still, health systems that embrace AI now may be positioned to lead in the future.

Latest news from @DeloitteHealth

Endnotes:

¹The benefits of the latest AI technologies for patients and clinicians, Harvard School of Medicine, August 30, 2024

²The growing nationwide radiologist shortage, Radiological Society of North America, March 4, 2025

³Is generative AI a reliable tool for medical self-diagnosis, July 4, 2023; Self-diagnosis through AI-enabled chatbot-based symptom checkers, American Medical Informatics Association, January 25, 2020

The executive’s participation in this blog is solely for educational purposes based on their knowledge of the subject and the views expressed by them are solely their own. This article should not be deemed or construed to be for the purpose of soliciting business for any of the companies mentioned, nor does Deloitte advocate or endorse the services or products provided by these companies.

This publication contains general information only and Deloitte is not, by means of this article, rendering accounting, business, financial, investment, legal, tax, or other professional advice or services. This article is not a substitute for such professional advice or services, nor should it be used as a basis for any decision or action that may affect your business. Before making any decision or taking any action that may affect your business, you should consult a qualified professional advisor.

Deloitte shall not be responsible for any loss sustained by any person who relies on this article.

Return to the Health Forward home page to discover more insights from our leaders.