In the United States, physician use of AI rose from roughly 38% to 81% over three years1. More than 1,500 AI-enabled devices (as of the end of Q1 2026)2 now hold FDA authorization. Patients have moved just as fast as approximately 72% of clinicians said patients arrive at consultations AI-informed3. But adoption should not be confused with scale.
Look closely at how doctors use AI, though, and the picture changes. Their most common uses are summarizing medical research and drafting notes and care plans. Only 17% use AI to help with diagnosis1. AI has moved quickly through the paperwork around medicine, where an error costs a few minutes. Its progress into diagnosis and treatment has been far slower, because an error there reaches the patient.
For global Health & Pharma* enterprises, that gap is where AI budgets stall. A model can impress in a pilot and still never make it into everyday clinical work. Every hospital and every regulator asks the same question before committing: does this model work on our patients and our equipment? The enterprise that answers it fastest will scale its AI first, and India’s Health & Pharma GCCs* are well placed to be the ones answering it.
*Health & Pharma includes four types of companies:
Healthcare:Â hospital and provider networks, payers, health services firms and health IT companies.
Pharma:Â companies that discover, develop, manufacture and commercialise drugs.
Biotech:Â companies developing biologics, including vaccines and cell and gene therapies.
Med Devices:Â companies making medical equipment and diagnostics, including software as a medical device (SaMD).
AI stalls after the pilot because hospitals and regulators want proof that a model works in the setting where it will be used, and that proof is slow and costly to produce.
Most models learn from data gathered at large, well-equipped hospitals. Move one to a community hospital with older scanners and patchier records, and its accuracy can fall. Human clinicians adjust to that kind of variation as a matter of routine, while models often do not. The industry calls this model-market fit. A model can pass every technical test and still disappoint the hospital that bought it.
Healthcare has no standard benchmarks to settle the question, and the results vendors publish tend to show their products at their best. Hospitals have responded by running their own checks, testing each model against their own past cases before they trust it. Every provider now carries its own validation burden, and every vendor faces a new test at every door.
Regulatory clearance leaves much of this open. About 97% of AI-enabled devices reach the US market through the FDA’s 510(k) route, which asks a manufacturer to show its device is substantially equivalent to one already on sale4. That establishes comparability with an earlier product. It says far less about how a model will perform on a particular hospital’s patients. Clearances are also bunched in a single specialty, with radiology accounting for about 76% of authorised AI devices2.
Doctors decide how far a model travels. In the AMA survey, 85% of physicians said they want to be consulted on, or responsible for, bringing AI into their practice1. They will ask for evidence from settings that look like their own.
Providers pay for AI when it produces a result they can measure, such as faster report turnaround or more patients screened. A hospital already short of staff has little appetite for a software licence that adds work, and any return has to be weighed against the cost of installing the tool and keeping it monitored.
The products winning adoption are priced as services that add capacity. One AI-native radiology platform now works with more than 2,000 hospitals. A blood-based early cancer test raised screening uptake by replacing an existing test that patients found less acceptable.
That buying pattern turns validation into a commercial question as well as a regulatory one. A vendor that cannot show results in a hospital’s own setting has little to sell there. An enterprise that can produce that evidence quickly, across many settings at once, reaches scale well ahead of one that validates hospital by hospital.
Clinical validation is the work of proving that an AI model performs safely and accurately where it will actually be used, on the patients and equipment it will meet there. It goes a step beyond technical validation, which tests a model against held-back data while it is being built.
Pharma and biotech companies face the same discipline with a different set of models. AI now supports trial design, patient selection, pharmacovigilance and manufacturing quality. When its output informs a regulatory decision, it has to earn credibility for that specific use.
The FDA’s first guidance on AI in drug development, issued in draft in January 2025, sets out a seven-step, risk-based framework for establishing a model’s credibility in a defined context of use5. The guidance also expects sponsors to keep monitoring a model’s performance across its life cycle5.
A device team proving an algorithm works in a new hospital and a pharma team proving a model can support a trial decision are meeting the same test.
Validation also has limits worth stating plainly. A model predicts outcomes; it cannot explain why they happen. Clinicians need that explanation before they act, which is why trust in how a model was tested matters as much as its accuracy score.
India’s Health & Pharma GCCs can bring together, in one location, the functions clinical validation depends on: biostatistics, regulatory, medical and technology. That combination is their real advantage, more than depth in any single function.
The base to build on is already large. In FY2026, 142 Health & Pharma companies ran GCCs in India, operating 229 GCC units and employing about 184,000 people6. Pharma and healthcare companies form the core, accounting for 147 GCC units and roughly 150,000 of the skilled talent already working in these units. Medical device makers add 58 GCC units and nearly 26,000 people, and Biotech companies another 24 GCC units with close to 8,000. A talent base of this size can staff validation work across every product line an enterprise runs.
The wider ecosystem points the same way. The Nasscom-Zinnov FY2026 landscape report found that centres built for delivery are increasingly holding global ownership7. It also found that 96% of GCCs set up since FY21 began with a product, R&D or engineering mandate6. A centre that already builds the models is a short step from owning the evidence behind them.
India’s clinical diversity adds a second advantage. Its hospitals run on a wide range of equipment and data quality, including the lower-resource settings many Western validation programmes leave out, which makes Indian data valuable for testing how well a model generalises. That advantage holds only if data governance and interoperability keep pace. It also has a limit. Outcomes vary with a population’s genetics, biology and social context, so each market still needs evidence from its own patients, and an FDA submission still needs data that reflects the US population.
The model already works. Several India-based and India-origin companies hold US FDA clearance for AI diagnostics that were developed and largely validated in India, and they sell into multiple countries. India-originated innovation can move into an ownership position when the validation behind it is credible.
A GCC earns ownership of clinical validation by holding accountability for the evidence that a model is safe to scale. That accountability can arrive two ways. Established centres earned theirs through years of consistent delivery. Newer centres sometimes receive it as a mandate from headquarters, and a mandate holds only if the centre can prove it deserves one. Four moves turn either route into lasting ownership.
Start where the idea starts. When an innovation originates in India, ownership can naturally stay there, along with the validation that proves it. Relocating a team that works on someone else’s idea rarely creates the same authority. The India-validated diagnostics now cleared by the FDA show where that path leads.
Build teams that can do the proving. Validation needs people who combine domain expertise in medicine or drug development with AI, data and product skills. Placing biostatisticians and clinical data managers alongside engineers means the question of how a model will be proven shapes how it gets built.
Work with the ecosystem around the centre. India has a large base of healthcare startups, many of them building exactly the tools a validation programme needs. Recruiting patients and securing ethics clearances for studies remains slow, and a shared clinical-study platform that handles both is where public support could make the biggest difference for GCC teams and startups alike.
Judge the centre by outcomes. The measure of ownership is enterprise results, such as a model cleared, a product adopted or a patient-impact target met. How much of the work physically sits in India matters far less. For device makers, accountability means answering for the clinical evidence behind FDA and CE submissions. For pharma and biotech companies, it means the credibility documentation behind drug and biologic filings and the pharmacovigilance that follows a launch. India’s centres work best as complementary global teams, each judged on what it delivers for the enterprise.
The enterprise gains one team accountable for whether its AI is safe to scale, and its India centre moves into decisions that shape clinical and regulatory strategy. Physician use of AI will keep climbing. How much of that use reaches patients directly depends on how fast enterprises can prove their models hold up in the real world. The centres that own that proof will decide which models scale.
At Zinnov, we have spent over two decades helping global enterprises navigate this exact evolution. Whether you are building a healthcare center from the ground up or scaling an existing hub to take on global AI accountability, our GCC Setup and Transformation advisory equips you with the blueprint to move from operational setup to scaled innovation.
Connect with our experts to build a future-ready healthcare GCC that owns the decisions that matter.
THANK YOU
FOR YOUR RESPONSE
