Healthcare systems across the UK, Europe, and North America are running into the same wall. Workforce numbers are not keeping up with demand, patient data sits in fragmented systems that don’t talk to each other, and clinical workflows still rely on manual steps that should have been automated a decade ago. The pressure is real, and the appetite for AI-assisted solutions is enormous. The catch is that off-the-shelf AI does not survive contact with a hospital.
Manu Agrawal is one of the engineers working on what does. She leads agentic AI initiatives at Oracle Health, focused on building systems that operate inside regulated healthcare environments without becoming a liability. Her earlier career was spent at Amazon Web Services, where she held senior engineering roles across Amazon Bedrock, CloudFront, and multimodal AI infrastructure.
From Hyperscale Cloud To The Hospital Floor
Agrawal is an enterprise engineer who deliberately moved into healthcare AI, and the move shaped how she thinks about the problem. At AWS, she was a founding member of an AWS service and owned the architecture for Amazon Bedrock Data Automation, the company’s first multimodal unstructured-to-structured data offering. That work taught her how to build systems that run at scale. Healthcare taught her what running at scale means when the system must be auditable, explainable, and overrideable by a clinician at any time.
She is also active in the broader AI community.
“In addition to enterprise leadership roles, I am actively involved in the broader AI and technology community through conference reviewing, technical publications, media features, workshops, and industry discussions focused on enterprise AI systems, responsible AI deployment, and governed autonomous systems,” she said.
More from Artificial Intelligence
- Intelligent Apps That Build Themselves Are Here – Traditional Software Engineering Is In Crisis
- Regulated SaaS Businesses Have 10 Weeks To Get Clarity On AI – Scail’s Risk and Value Scorecard Fixes That
- How AI Expansion Is Increasing Silver Demand
- AI Gave Everyone A Lawyer – Now The Courts Are Drowning
- Behind The Anthropic And Vatican Partnership: How An AI Startup Is Pushing For Ethics
- The Rise Of The Executive AI Twin – Meet The Leaders Who Have Outsourced Themselves
- The UK Just Admitted It Lost The AI Chip Race – Here Is Its Plan B
- How Much Data Do UK Internet Users Actually Give Tech And AI? New Research Tells The Shocking Truth
The Unglamorous Work Behind Agentic AI
At Oracle Health, Agrawal works on what the industry now calls agentic AI: systems that can take actions within clinical and operational workflows without operating as black boxes. The applied problems are practical and recognisable to anyone running a healthcare organisation. Care gap closure, where the system identifies patients whose follow-up has fallen through the cracks. Clinical workflow automation, where AI handles the repetitive administrative work that consumes clinician time. Cohort discovery for medical research. Trial intelligence and evidence generation.
These initiatives support healthcare transformation efforts involving large healthcare providers, public-sector healthcare organisations, and healthcare systems operating across multiple countries and regions. The work is global in scope, but the operational problems are local: every hospital has its own data systems, its own workflows, and its own constraints. AI that cannot adapt to that variance does not get deployed.
Why Most Healthcare AI Fails The Trust Test
Most healthcare AI work splits into two camps. The first builds clever models that struggle to leave the research stage. The second builds enterprise software that uses AI as a label rather than as a functioning component. Agrawal works in a third position, which is rarer and harder to staff: someone with a distributed-systems background to ship at production scale and the discipline to build for regulated environments from the start.
Her framing of the problem is unusually consistent. Governance, orchestration, and human oversight are not policy concerns to be addressed after the system is built. They are architectural requirements that determine whether the system can be built at all. In a sector where most AI conversations still revolve around model capability, that is a noticeably different conversation.
It is also the conversation healthcare actually needs. Workforce shortages will not be solved by clever models alone. They will be solved by AI systems that clinicians trust enough to delegate to, and that hospital administrators trust enough to deploy at scale. The bar for trust in healthcare is higher than in any other enterprise sector, which is why so few systems clear it.
The handful of engineers building toward that bar are quietly defining what the next generation of healthcare technology will look like. Agrawal is one of them, and the playbook she is developing is one that the UK and European healthcare sectors will need as much as the American market it is currently being built for.
