Insights by Bharat Sutariya, MD, SVP and Chief Health Officer, Oracle Health
Key Points
- Oracle is making a bold bet that “bolt-on AI” will hit a ceiling, and that the real transformation requires rebuilding the healthcare stack with AI embedded at the foundation.
- The next leap is moving from systems of record to systems of orchestration, where AI listens for clinical intent and queues orders, referrals, and prior authorizations.
- Near-term value comes from reducing high-volume friction, while governance and human-in-the-loop guardrails remain essential as AI moves closer to clinical decision support.
“AI needs to do the work.”
That single line captures the urgency and ambition in this conversation with Dr. Bharat Sutariya. His core argument is not that healthcare needs more AI features. It needs a different architecture. One where AI is not bolted onto legacy workflows but embedded into the foundational layer so the system can orchestrate work across the care journey.
Dr. Sutariya’s perspective is shaped by living through every era of modern health IT. He is an emergency physician by training with 25-plus years at the intersection of healthcare and technology, including leadership roles at Detroit Medical Center, 17 years at Cerner, a stint at Deloitte, and now as Senior Vice President and Chief Health Officer at Oracle Health. He has seen healthcare move from paper to EHR documentation overload. His “origin story” into technology is simple and relatable: impatience with things that do not work, paired with a relentless drive to improve care at scale.
What makes this episode especially relevant is that it confronts the question every CIO, CMIO, and CEO is wrestling with right now: are we heading toward incremental productivity gains, or toward a fundamentally different operating model for care delivery?
Dr. Sutariya believes the answer depends on whether we keep bolting AI onto legacy systems or build platforms that treat AI as the new core of the workflow.
Listen to the full conversation
Why “Bolt-On AI” Is Only the Beginning
One of the most useful parts of this episode is the way Dr. Sutariya reframes what people blame on the EHR.
As he explained to The Big Unlock podcast host, Ritu M. Uberoy, “the EHR itself is not the only culprit. A large portion of the burden clinicians feel stems from the compounded weight of regulatory compliance, medical-legal requirements, expanded evidence requirements, and administrative demands layered onto the digital workflow over decades. The EHR became the container for all of it, so the frustration is often directed at the EHR.”
That is why he believes AI matters, but also why he believes the usual pattern of adding tools on top of legacy infrastructure will only take the industry so far.
He calls out what has become the industry norm, keeping the foundational EHR and bolt AI on top. He acknowledges that both “bolt-on” and “rebuilt” approaches can show early success, especially because the industry is still “scraping the surface” of what AI can do. But he predicts differentiation will come when AI moves from documentation and isolated agents to orchestration across workflows.
Oracle’s bet, as he told Ritu, is different. Rather than treating AI as an add-on, Oracle is rebuilding the healthcare tech ecosystem across providers, payers, and life sciences using the full Oracle stack, from database and cloud infrastructure through the AI layer to modern applications. In his words, Oracle is embedding AI into the EHR, or even more radically, embedding the EHR inside the AI.
That is not just a product story. It is a sequencing story. It says the future is not “a smarter form.” The future is a workflow engine that can interpret intent, coordinate tasks, and reduce the load clinicians carry every day.
The Shift From Documentation to Orchestration Is Already Underway
When Dr. Sutariya talks about near-term value creation, he is very specific: “reduce high-volume friction,” or the repetitive work that drives burnout. Documentation, chart review, ordering, and follow-up tasks. He argues that AI agents deployed with full chart context can meaningfully reduce burden, improve satisfaction, reduce pajama time, and improve patient interaction because clinicians are not hiding behind keyboards.
He also makes an important point about outcomes. In addition to measuring process metrics like time saved, he says Oracle is increasingly tracking clinical outcomes, financial outcomes, patient experience, and whether clinicians can operate more efficiently without sacrificing quality. That is an important evolution for the industry, because the “time saved per note” story is not enough to justify long-term platform modernization.
Then he explains what “orchestration” looks like in practice, and this is where the conversation gets concrete.
Ambient documentation is not the endpoint. It is the foundation.
As Oracle’s ambient agent listens to the clinician-patient conversation and drafts the note, it also extracts clinical intent and begins queuing actions:
- orders mentioned by the clinician
- prescription renewals
- referrals to other clinicians
- and, when appropriate, prior authorization workflows
He describes a scenario where the agent hears a clinician discuss a knee replacement. The system identifies payer requirements, retrieves prior authorization criteria, gathers relevant chart information, fills the required documentation, and presents it for clinician review or routes it to the appropriate queue for completion.
The key distinction he is drawing is that the AI is not “making medical decisions.” It is capturing clinician intent and automating downstream administrative work that normally slows care, creates delays, and drains staff capacity.
This is what he means by AI doing the work.
It is also what he means by moving from a system of record to a system of orchestration. A system that does not simply store what happened, but helps move the care journey forward.
Guardrails, Transparency, and Sequencing Are What Make This Safe
Dr. Sutariya is clear that healthcare has a different error tolerance than most industries. The “parts per million error rate” that might be acceptable elsewhere is not acceptable in patient care. Therefore, the path to AI-native orchestration cannot be reckless.
His answer combines governance, safety sequencing, and transparency.
First, he emphasizes starting with lower-risk areas, such as operational and administrative workflows. There is significant efficiency and quality improvement that can be achieved before AI moves into higher-risk clinical domains.
Second, he argues that you need trusted platforms with appropriate guardrails and governance. This is partly why he advocates moving away from a fragmented ecosystem of dozens of AI startups, each requiring data extraction and creating new cybersecurity and privacy burdens. He suggests health systems should accept that AI will be a foundational component and choose a partner or a small set of partners that can provide infrastructure, governance, and reusable services across many use cases.
Third, he describes how Oracle is thinking about transparency in assistive clinical scenarios. He gives a clear example, telling Ritu that “AI-generated chart summaries should not be opaque.” Clinicians should be able to see the source of each critical fact. He describes a workflow where clinicians can hover over a key summary element to see its source, and, with a single click, open the underlying document with the relevant text highlighted. That approach preserves clinician trust and reduces hallucination risk by making provenance visible.
This is a key theme: as AI becomes more capable, the difference between safe acceleration and unsafe automation will be traceability, explainability, and control.
Dr. Sutariya’s viewpoint is that the right platform can embed those safeguards at the infrastructure level, rather than forcing each use case to reinvent them.
The Next Chapter Is Orchestration — Not More Documentation Tools
Dr. Sutariya predicts that within a year, the conversation will move away from documentation efficiency and toward orchestration. AI should not create work. It should do work.
He also makes the North Star explicit. None of this matters unless clinicians and patients feel they get time back, are doing fewer repetitive tasks, and experience a safer path toward better care. Health systems will only win if the care delivery engine improves in a way that clinicians can sustain and patients can feel.
In his framing, AI-native orchestration is not a feature. It is the path to a connected, intelligent ecosystem where intent becomes action and administrative work no longer consumes the clinical day.
The Takeaway
Dr. Bharat Sutariya’s message is both ambitious and practical. “Bolt-on AI” can deliver early wins, but healthcare’s real transformation requires AI-native orchestration built into the foundational platform. Dr. Sutariya’s prediction is that the industry will quickly move from “documentation efficiency” conversations to “orchestration” conversations, and that winners will be the health systems and platforms that sequence safely, maintain transparency, and deliver what clinicians and patients actually want: time back, fewer repetitive tasks, and a more connected care experience.
Sitting at the intersection of emergency medicine, decades of EHR evolution, and AI-native platform strategy, Dr. Sutariya’s unique insights are especially valuable:
- EHR “burden” is compounded by decades of compliance and administrative layering, and AI is an opportunity to unwind friction rather than add another layer.
- Bolt-on AI will show early success, but orchestration will differentiate platforms once health systems demand end-to-end workflow impact.
- Ambient is the beginning, not the end. The real value is when AI listens for clinician intent and automates follow-on tasks.
- Orchestration means connecting the conversation to action: queuing orders, referrals, and prior authorization workflows while preserving human control.
- Decision fatigue is real. Health systems should choose trusted AI partners and infrastructure rather than bolting on dozens of point solutions.
- Transparency and provenance are essential as AI moves closer to clinical workflows, and must be designed into the platform, not patched on.