Adopting and scaling AI: Reflections for executive leaders in healthcare

• Define your long-term AI vision before you scale your first pilot
• Data quality and interoperability are non-negotiable foundations
• Trust is built through transparency, co-design, and real-world results

While the promise of integrating AI technologies in healthcare is vast, its true value will be borne out by seamlessly scaling AI into daily operations. After a successful pilot, healthcare leaders must create an ecosystem to support AI at scale. This is a challenge that requires careful alignment between corporate strategy, digital infrastructure, risk governance, and workforce evolution.

Last week, we published a discussion on overcoming organizational challenges to AI in healthcare with one of the industry's leading voices, Shweta Maniar, Global Director for Healthcare and Life Sciences at Google. This week, we continue our discussion, shifting the focus to how healthcare leaders should consider the challenges of scaling AI and the strategic imperatives that can bring truly transformational change to healthcare organizations.

HT: Why is it crucial for healthcare organizations to plan for scaling AI from the beginning?

Shweta Maniar: You don’t have to start at scale, but you do need to understand what scale looks like for your organization. Too often, teams launch a successful pilot, maybe in one clinic or one department, but they hit a wall when they try to expand. Why? Because scale wasn’t part of the original design.

What’s your ultimate vision? Are you aiming to support thousands of clinicians, or integrate with dozens of legacy systems? If so, your pilot needs to be built on foundations that can grow with you. Think about infrastructure, data flow, governance, training, and even the budget needed for long-term support. It’s the difference between building a prototype and building a platform.

Having this mindset from the beginning doesn’t mean over-engineering, but it does mean being intentional. Even if you're starting small, make sure the decisions you make today won’t limit your progress tomorrow. 

HT: How does investing in data infrastructure contribute to successful large-scale AI applications?

Shweta Maniar: It’s everything. Your AI is only as powerful as the data behind it. In healthcare, that data is often siloed, inconsistent, or unstructured. If you're serious about scale, you have to get serious about your data architecture.

That means building pipelines that can handle structured and unstructured data: lab results, imaging, clinical notes, even wearable device data. It means ensuring interoperability between systems, so data can flow securely and efficiently across platforms and departments. It also means making data quality a priority from day one.

I always use the baking analogy: if you have a world-class oven but your ingredients are mislabeled or expired, you’re not going to get a good result. That’s exactly how integrating AI technologies into healthcare works. You need clean, well-organized, accessible data to drive consistent and reliable outcomes. 

HT: What steps should organizations take to build trust in AI among non-technical users and stakeholders?

Shweta Maniar: Trust isn’t a nice-to-have; it’s foundational. Especially in healthcare, where people’s lives and well-being are at stake. If clinicians, staff, or patients don’t understand how a tool works or don’t believe in its value, they won’t use it, no matter how accurate it is.

First, prioritize explainability. AI tools should be transparent about how they arrive at decisions, especially in clinical settings. If an algorithm tells a radiologist to prioritize certain scans, they need to understand why. That transparency builds confidence and makes adoption of AI in healthcare easier.

Second, provide tailored training. Not everyone needs to become a data scientist, but everyone should feel equipped to use the tool in a meaningful way. Furthermore, the training should reflect real-world scenarios, not just theory.

Third, and this is crucial, involve end users early in the design process. Co-create with them. When people help shape the solution, they’re more likely to trust it and champion it. That sense of ownership can be more powerful than any training manual. 

HT: How can organizations effectively demonstrate the ROI of AI adoptionI in healthcare?

Shweta Maniar: The return on investment has to be clear, not just in impact on the organization but in terms of impact on the patients as well. Clinical outcomes, operational efficiency, patient experience, and staff satisfaction all count, and they all contribute to long-term success.

Let’s say a hospital implements an AI scheduling tool. The financial ROI might include reduced no-show rates and more efficient use of clinical time. However, the operational ROI could include fewer delays, better workload distribution, and less clinician frustration. From a patient perspective, the ROI might show up as shorter wait times and better communication.

It’s not one-size-fits-all. The key is to align your metrics with the outcomes your stakeholders care about. Be sure to capture both quantitative and qualitative data - stories and satisfaction scores matter just as much as percentages and dollar signs.

HT: What strategies should be employed to ensure AI enhances full workflows rather than isolated tasks?

Shweta Maniar: One of the most common mistakes I see is treating AI as a patch rather than a pathway. Organizations implement a tool that solves one problem, but it doesn’t connect to the broader workflow. That’s a missed opportunity.

Start with process mapping. Understand the full journey, from the first patient interaction to the final outcome. Identify bottlenecks, handoffs, redundancies. Then ask: where can AI provide meaningful, integrated support?

The goal is end-to-end improvement, not just task automation. For example, if an AI tool predicts hospital readmission risk, what’s in place to act on that prediction? Is there a care coordinator looped in? Is there an intervention pathway built into the system?

You also need to plan for iteration. AI isn’t a “set it and forget it” solution. As workflows evolve and data improves, your models and processes need to adapt too. That ongoing refinement is what drives real, lasting impact.

HT: Thank you so much for your time today. One last question: What resources can you recommend to executive leaders for scaling AI within a healthcare system?

Shweta Maniar: The most successful AI transformations start with a few high-impact projects, not a "boil the ocean" strategy. This approach allows your teams to learn, build confidence, and see the value firsthand, making them your best champions for scaling. There are some great resources for those who want to find out more:

• Healthcare Executive Podcast by the American College of Healthcare     Executives: This podcast offers short, digestible episodes on a variety of leadership topics, including digital transformation and AI.

• CareTalk: Healthcare. Unfiltered: This podcast offers candid, often unfiltered conversations about the business of healthcare, technology, and policy.

• The Digital Doctor: Hope, Hype, and Harm at the Dawn of Medicine's Computer Age by Robert Wachter, MD: This book is a must-read. Dr. Wachter provides a grounded, often humorous, look at the challenges and opportunities of technology in medicine. It's not just about AI, but it offers a vital perspective on the human and organizational factors that determine whether a new technology sinks or swims. It's a fantastic resource for understanding the cultural shifts required for success.

• Harvard Medical School's "AI in Health Care" Program: An online program for helping health care leaders design, pitch, and implement AI-driven solutions for their organizations.

• McKinsey's Scaling National e-health Report:This useful report shares strategic actions that successfully spurred adoption and regular use of e-health solutions in countries around the world and offers suggestions about how other countries can combine and implement similar programs.