70% of Healthcare Organisations Use AI in 2026, Says NVIDIA-ROI, AI Agents and Open Source

NVIDIA: How 70% of Healthcare Organisations Are Using AI in 2026

Healthcare AI has moved from pilots to production. NVIDIA's latest "State of AI in Healthcare and Life Sciences" survey of 600+ professionals shows broad adoption, clear ROI, and growing budgets across clinical, research and operational work.

The headline: 70% of organisations now use AI (up from 63% in 2025), and 69% use generative AI and large language models. The action is spread across imaging, documentation, drug discovery, logistics and administrative streamlining.

Adoption snapshot

• Active AI use: 70%. GenAI/LLMs: 69% (up from 54%).
• By segment: Digital health 78%, pharma/biotech 74%, medtech 70%, payers/providers 56% (up 13 points year over year).
• Foundational analytics: 65% use data analytics/data science; 51% use predictive analytics.
• Top clinical/ops use cases: clinical decision support (42%), medical imaging (38%), administrative workflow optimisation (38%).

"Over the next 12 to 18 months, the most visible and scalable impact of AI will come from logistics and administrative streamlining," says John Nosta, President of NostaLab. "That's where adoption curves are already steep - scheduling, documentation, coding, utilisation management and care coordination."

Where ROI is showing up

AI pays when you target specific, well-defined problems. That's consistent across the data.

• Medtech: 57% report ROI from AI in medical imaging.
• Pharma/biotech: 46% report ROI in drug discovery and development.
• Digital health: virtual health assistants and chatbots rank as top ROI drivers.
• Payers/providers: administrative task automation and workflow optimisation lead.

Across the industry, the strongest ROI appears in three buckets: medical imaging, workflow optimisation and NLP for clinical documentation.

"Scaling generative AI in healthcare starts with focusing on real clinical and operational problems, rather than the technology itself," says Dr. Annabelle Painter, Clinical AI Strategy Lead at Visiba U.K. "The organisations seeing impact are those that embed AI into existing workflows instead of layering AI on top as a separate tool. The ones that win fund and prioritise evaluation so AI improves safety, quality and patient care over time."

Agentic AI and open-source momentum

Agentic AI - systems that can reason and complete tasks - is moving from slideware to use. 47% say they're actively using or assessing AI agents. 22% have deployed them; another 19% plan deployment within a year.

• Top agentic AI use cases: knowledge management and retrieval (46%), literature review and analysis (38%), internal process optimisation (37%).

In pharma and biotech, 55% use agentic AI for literature review, and nearly half apply it to drug discovery and biomarker identification.

Open source is central to this push. 82% say open-source models and software are moderately to extremely important to their strategy, enabling fine-tuning for specialised clinical and research tasks. Hybrid computing is rising too: 43% now use hybrid infrastructure for AI projects (up from 35%).

Budgets, infrastructure and scale

Investment is following results. 85% expect AI budgets to increase in 2026; nearly half expect growth above 10%. Priorities are shifting from experiments to scale: 47% will focus on optimising AI workflows and production cycles (up from 34%).

Infrastructure is catching up. 34% plan to build or expand AI infrastructure in 2026 (up from 24%). On business outcomes, 85% of management respondents report AI increased annual revenue and 80% report reduced annual costs. 44% saw revenue rise by more than 10% - with small companies most likely to report double-digit gains. On costs, 35% overall (and 44% of small companies) cut more than 10%.

Barriers you should plan for

• Smaller organisations: budget constraints (40%), insufficient data for training (33%).
• Larger enterprises: data privacy and security concerns (39%), regulatory and ethical issues (37%).
• Agentic AI: 40% say compliance requirements strongly shape implementation strategies (HIPAA, FDA approvals, GDPR).

Helpful references: HIPAA Privacy Rule overview and the FDA's guidance on AI/ML Software as a Medical Device.

A practical playbook for healthcare leaders

• Pick 1-3 high-ROI use cases per line of service. Examples: medical imaging triage/QA, clinical documentation and coding support, utilisation management and prior auth, care coordination and scheduling optimisation, genomic variant calling and report drafting.
• Embed into existing workflows. Integrate with EHR/RIS/LIS and messaging systems. Avoid standalone tools that add clicks. Define clear success metrics (report turnaround time, denial rates, throughput, readmissions, time-to-first-decision).
• Stand up rigorous evaluation. Prospective A/Bs where possible. Track safety, bias, drift and user satisfaction. Publish dashboards. Make "evaluation" a funded, ongoing function - not a one-off project gate.
• Data stewardship first. De-identify where feasible, lock down PHI access, implement audit trails and role-based controls. Establish data retention and vendor data-use terms up front.
• Choose the right model strategy. Use open models for discovery and internal research where flexibility matters; use vetted proprietary systems for clinical deployment where validation, integration and accountability are mandatory.
• Build a pragmatic architecture. Expect hybrid compute (on-prem GPU + cloud burst). Containerise workloads, add MLOps/LMMOps, and monitor latency, cost and output quality. Align GPU capacity with your highest-value pipelines.
• Governance and regulatory path. Name a clinical safety owner, define RACI, document known risks, mitigation and human-in-the-loop checkpoints. Map features to applicable regulatory pathways early (including change-control for learning systems).
• Agentic AI with guardrails. Start with bounded tasks (knowledge retrieval, literature review). Constrain tools and data scopes. Require human sign-off for actions that affect care or compliance.
• Upskill your teams. Train clinicians, informatics and ops leaders on prompt design, review workflows and exception handling. Require vendors to provide audit logs, model cards and support for local evaluation sets.

What this means for your organisation

AI has become part of standard operations across healthcare and life sciences. The gains are clearest when you focus on narrow, valuable workloads and wire them into the way clinicians, scientists and administrators already work.

With adoption high and budgets growing, the advantage goes to teams that execute: pick the right use cases, integrate deeply, measure relentlessly and keep governance tight. If you need practical how-tos and training, see AI for Healthcare and explore AI Agents & Automation for admin and ops use cases.