AI Models Aren’t Medicine: Why Connected Care Wins in Healthcare

Hospitals built up their digital infrastructure piecemeal over decades, adding specialised systems to solve isolated operational problems rather than designing a connected ecosystem from the outset. However, this slow mix of old and new tools causes data blocks and connection problems.

Each new platform speaks its own digital language, and clinical information ends up scattered across electronic health records, laboratory systems, and radiology platforms that were never built to interoperate.

But Surjeet Thakur, Chief Information Officer at Rajagiri Hospital in Kochi, Kerala, wants to dismantle a common assumption: that data fragmentation is fundamentally a technology problem. 

He identifies governance, not technology, as the true obstacle. Even where interoperability standards such as FHIR (Fast Healthcare Interoperability Resources) exist, they only work when hospitals also standardise their clinical workflows and align their data definitions across vendors.

"Successful interoperability requires standardised clinical workflows, consistent data definitions, and alignment across vendors. Without these foundational elements, interoperability becomes an integration exercise rather than a transformation initiative," he tells AIM

Rather than waiting for every legacy system to become fully standards-compliant, the hospital has pursued an integration-first strategy built around middleware, APIs, and interface engines. 

This allows clinicians to work from a unified view of the patient without ripping out and replacing existing infrastructure. Thakur frames this as evolutionary change.

Measuring Value Beyond the Balance Sheet

Quantifying the return on AI investment in healthcare is notoriously difficult. According to a Qventus report, just 4% of healthcare organisations have achieved scaled AI implementation with measurable outcomes. 

Thakur argues that traditional financial metrics miss the point entirely. Hospitals like Rajagiri instead use a balanced scorecard that tracks reduced length of stay, lower readmission rates, faster clinical turnaround, improved diagnostic accuracy, and greater clinician productivity. 

Financial benefit, he suggests, follows naturally once clinical decision-making improves.

AI Tackles Alert Fatigue

This same patient-first logic shapes how the hospital tackles alert fatigue, a growing concern as AI systems generate ever more notifications for already overstretched clinical staff. Thakur insists that flooding physicians with alerts does nothing to improve care; instead, systems must intelligently filter information based on clinical context and urgency.

"The objective is intelligent prioritisation rather than notification overload. AI should understand clinical context, patient acuity, speciality-specific workflows, and urgency before generating recommendations. We focus on delivering high-value, actionable insights integrated seamlessly into existing clinical workflows."

For Thakur, every technology decision ultimately returns to a single test: does it reduce the burden on clinicians, or add to it? Automation that creates additional documentation or duplicate data entry, he argues, has failed regardless of its sophistication. 

This same discipline extends to how Rajagiri Hospital approaches privacy, where governance principles such as encryption, access controls, and data minimisation sit alongside emerging techniques like federated learning to enable collaboration without compromising patient confidentiality.

Three Planes of Transformation

Sridhar Turaga, SVP of the AI Practice at healthtech and consultancy firm CitiusTech, maps how AI is reshaping healthcare across three simultaneous planes. The first is assistance, where tools such as ambient listening and nurse-support agents ease the cognitive and administrative load on clinical staff.

"AI in healthcare is evolving in three planes, all in parallel: assistance to reduce cognitive burden and ease policy adherence for doctors, nurses, and care teams, giving back their time and energy for patient care and experience," he notes.

The second plane, actionable intelligence, focuses on automating operational processes such as claims processing, revenue cycle management, and supply chain logistics, driving measurable improvements in cost and cycle time. The third and most ambitious plane, which Turaga calls ‘agent-first design’, moves beyond software assisting people towards autonomous agents orchestrating entire processes across people, systems, and knowledge assets.

Turaga argues that healthcare's complexity sets it apart from almost every other consumer-facing industry. 

"Healthcare has a unique characteristic unlike any other consumer-centric industry, as any patient's care journey is fragmented across multiple entities, multiple stakeholders, and multiple objectives, with no unified view of the entire history, which makes technology and automation much harder than in other industries," he explains.

Despite this complexity, Turaga sees genuine cause for optimism in the evolution of AI. Foundation models are naturally suited to working with the messy, unstructured data that healthcare generates, while agentic AI is built specifically to coordinate across diverse actors and systems—precisely the capability healthcare has lacked until now.

"Agentic AI is natively born to orchestrate across diverse actors, and foundation models are natively born to work with fragmented and unstructured data, creating a new promise of automation and intelligence to reimagine healthcare and take healthcare operations, experience and services to a new level."

A Shared Destination

Looking to the future, Thakur is clear that the CIO's role is not to chase technological sophistication for its own sake, but to ensure that AI genuinely improves outcomes across a fragmented, multi-provider care journey.

"AI should help create connected, intelligent, and patient-centric healthcare ecosystems where information flows seamlessly, and care teams collaborate effectively," he notes.

Both Thakur and Turaga see fragmentation as healthcare's defining structural challenge, and view AI not as a replacement for clinical judgement but as a means of restoring time, context and coordination to an overburdened system.