AiThority Interview With Pramila Srinivasan, Ph.D., CEO, CharmHealth

The global medical industry is undergoing rapid transformation with the introduction of artificial intelligence, with AI now redefining diagnostics,  drug discovery and patient care. Pramila Srinivasan, Ph.D., CEO, CharmHealth weighs in with some observations:

________

Hi Pramila, tell us a little about CharmHealth and the journey so far? We’d love to know about your latest MCP server enhancement and how it benefits end users.

In speaking with clinicians and developers, I consistently hear that AI adoption is slow not because of lack of interest but because of integration complexity. Many hospitals want to deploy AI, but connecting those systems to clinical data in a safe and reliable way has proven harder than expected.

The model context protocol (MCP) server helps address that challenge. It allows AI applications to interact with electronic health record (EHR) data using natural language rather than requiring developers to build custom connectors for every use case.

I have seen how much time and effort goes into mastering API documentation and stitching systems together. With that complexity removed, teams can focus on solving real clinical problems.

For clinicians, the benefit is practical. AI can help with scheduling, summarizing charts, supporting documentation, or surfacing potential drug interactions — all within existing clinical workflows. The goal is to make AI usable in everyday care settings so clinicians can spend more time focused on patients rather than software.

Also Read -> Artificial Intelligence (AI) and The Future of Medical Care

Why is an MCP server more useful for driving optimized AI adoption? What points would you share with end users deploying AI-backed systems in their organizations and hospitals?

I have seen too many healthcare organizations move quickly to pilot AI tools, and then they realize later that the real challenge is with coordination. I have found that an MCP server can be valuable because it creates a more structured way for AI systems to access data and follow shared standards as well as to operate within clinical workflows.

We have all seen AI adoption break down when tools are deployed in isolation. Clinicians do not have time for disconnected systems, and IT teams cannot manage multiple models pulling from inconsistent data sources. I believe an orchestration layer can help ensure AI is pulling from trusted data and functioning across systems in a way that feels integrated rather than layered.

When I speak with hospital leaders about deploying AI-backed systems, I encourage them to start with workflow. Adoption will stall if AI does not fit naturally into how care teams already work. I also stress the importance of setting up standards and policies from day one. Audit trails, managerial oversight, and strong security are not optional in healthcare. It is important that the systems you put in place today leave room to grow so teams are not forced to start over as their AI use expands.

What trends are impacting EHR adoption, and how are providers finalizing product roadmaps based on these trends?

The most significant shift I am seeing right now is that providers are no longer asking whether to adopt AI — they see it as inevitable. But they are asking how to do it responsibly and in a way that works with how they work. This is a meaningful change from just a few years ago. It’s sort of a macro trend.

Diving a bit deeper, the administrative burden on clinicians that I mentioned earlier is now at a breaking point. Physicians are spending upwards of two hours a day on documentation alone. That’s time away from patients, and it contributes to burnout levels the industry just can’t sustain. EHRs can be an effective part of the solution, but only if they are designed in a way that doesn’t become another source of tension; they can’t give clinicians yet another thing to do that just looks different than what they had before.

At the same time, we are watching interoperability requirements mature. Regulatory pressure around data access and information blocking is pushing the industry toward a more open, standards-based infrastructure. That is actually a good thing. It creates the conditions for AI to function properly because AI can only be as good as the data it can access in a consistent, governed way.

What I hear consistently from providers as they build out their roadmaps is that they want to use AI, and they want to have modern tools that make life easier, but they are getting stuck. In the search for new technology, many practices have piloted three or four point solutions over the past few years — a standalone AI scribe, a patient engagement app, and so forth. This is great in terms of a willingness to move forward, but it creates some new problems. Now they have disconnected data, redundant vendor relationships, and staff who simply stop using tools that are not effectively talking to each other.

So the roadmap conversation has shifted from adding more and more capabilities to how do we integrate and simplify what we already have so that we can get the most out of it? When they figure out that missing piece of the puzzle, they figure out how to get work done in a manner that is helpful to both clinicians and patients — and again, this is where MCP can play a vital role. This is the direction I believe the market is moving.

Can you highlight some of the most innovative ways in which AI has enabled clinicians and hospitals with better administrative and other care outcomes?

The examples that resonate most with me are the ones that seem almost quiet in their impact until you take a step back and realize how much has changed.

Take nurse triage. Traditionally, a nurse conducts patient intake and then manually enters every piece of information from that patient conversation into an EHR. Allergies, current medications, diagnoses, family history, social history — screens worth of data manually entered. No one likes to do that. It’s repetitive, error-prone, and it takes time away from actual patient care. But until recently, it’s been necessary. Now with AI listening, extracting and recording all of that information in the moment, the nurse can review and confirm a structured summary rather than create a summary from scratch. This solves not only the efficiency problem but reduces the cognitive load that can lead to mistakes.

On the provider side, we have seen 95% note accuracy through AI-assisted documentation, with a 30-50% reduction in administrative burden. Now these are early results, but they are very compelling. When a provider walks into an exam room after already reviewing an AI-generated patient summary, with active diagnoses, medication history, recent lab trends, and flagged interactions front and center, the doctor is more prepared for the exam. The visit is more focused and helpful — and patients notice!

Where I think people underestimate AI’s potential is in the decision support layer. While ambient AI that transcribes is undoubtedly useful, AI that reasons across the full clinical context is where outcomes improve. Think about how AI can surface a potential drug interaction the provider might not have caught or its ability to identify a lab trend that might deserve follow-up testing. The key is that all decision-making remains with the doctor. He or she has the final say. Human judgment and expertise matter. AI presents the reasoning, but the clinician reviews and acts. This is non-negotiable.

On the operational side, the impact is perhaps not as visible but still really important. Scheduling, task management, lab result routing, billing support — each of these areas can improve with the strategic application of AI. And this has a direct effect on the quality of patient interactions.

Five quick thoughts on the future of AI and medical care before we wrap up?

1. AI will not replace the clinician. It will redefine how they spend their time and how they make decisions about patient care. The providers who thrive will be the ones who allow AI to work alongside them, complementing and assisting them, not the ones who resist it. Physicians will go back to doing what they do best. Connecting with patients and caring for them.

2. Infrastructure matters more than features. I see a lot of excitement about individual AI capabilities, which is warranted (I am excited too), but sustainable adoption depends on whether the underlying systems are interoperable, secure and built to last.

3. Trust is earned through transparency. Clinicians need to see and understand the reasoning behind an AI’s recommendation, not just the output. This is the only way a provider can accept responsibility for actions taken based on that recommendation. That’s how we maintain systems of accountability.

4. We have not really touched on this, but equity has to be a design principle. AI trained on incomplete or biased data will produce incomplete or biased recommendations. As an industry, we must be rigorous in our design.

5. The best outcome for AI in healthcare is one you barely notice. When the documentation happens, when the relevant information surfaces, when the workflow fits the way care teams actually work, AI is doing its job. The goal was never a flashy demo. It’s a clinician who got home on time and a patient who felt genuinely heard and cared for by their provider and staff.

Also Read: ​​The Infrastructure War Behind the AI Boom

[To share your insights with us, please write to psen@itechseries.com]