Health care providers will face compounding structural pressures over the next decade: fast-rising demand, constrained supply of health care professionals, and no easy way to change prices. Labor costs, already 50% to 70% of provider cost structures, continue to increase while access constraints worsen and margins compress.
In this difficult environment, incremental efficiency gains will not be sufficient; providers need step-change productivity improvements that expand capacity without proportional labor growth. In this context, AI is more than a technology; it enables the CEO to rebuild how care is delivered.
Why Now?
The pressures are particularly stark in the US. An aging population and the rising burden of chronic disease mean demand for health care professionals will be 14% higher than supply by 2036, according to industry and official projections. Despite this fast-rising demand, the economic model is deteriorating: by 2033, reimbursements will have grown by 27%, but hospital wage bills will be higher by 44%, while health care costs are growing around 5.8% annually, three times the 1.8% annual GDP growth.
Without fundamental change, waiting times will lengthen and margins will fall significantly.
Stay ahead with BCG insights on health care payers, providers, systems & services
AI as a Capacity Multiplier
Most AI efforts in health care have struggled to scale, not because of technology limitations but because providers have layered AI onto workflows that were never designed for real-time intelligence.
Recent advances in generative AI and autonomous AI agents enable systems that reason across unstructured data and act within workflows in real time. This allows AI-first providers to start from a different premise; value is unlocked by reimagining the way work is done today and creating new workflows from the ground up so that:
- AI handles information synthesis, coordination, and routine decisions.
- Humans focus on exceptions, judgment, and complex care.
- Work is orchestrated dynamically across the patient journey rather than confined to silos.
In terms of workflow, this means:
- Patients are triaged and routed to the most appropriate and efficient site of care.
- Clinical context is prepared before encounters, giving clinicians access to history, risks, and likely pathways.
- Documentation and coding are automated in real time.
- Financial and clinical risks are discovered early and addressed proactively.
This makes better use of health care’s scarcest resource: clinical talent. The result is a fundamentally better productivity equation: higher throughput, lower unit cost, reduced administrative burden, and improved patient experience.
Implementing AI at Scale
This rebuilding will succeed only if effort is distributed in the following way:
- 10% of effort on algorithms and use-case-specific models. Although this is critical, most providers put too much focus here, overthinking vendors and AI models.
20% of effort on building a scalable, modern data and technology infrastructure. Providers often perceive this as daunting and, as a result, defer it to platform vendors. In reality, they cannot outsource these processes entirely. Infrastructure must be built progressively alongside use cases, with providers owning the sequencing of capability building and the data orchestration layer at the center.
This foundation must deliver unified, real-time data access, supported by strong controls across cybersecurity, Responsible AI, identity and access management, and AI model validation. In the health care environment, this foundation is essential for safe deployment.
- 70% of effort on workflow redesign, role redefinition, and adoption. This is where the majority of value is realized—by embedding AI into how work gets done and by shifting human roles from execution to orchestration. This requires operational and cultural redesign at scale, driving change management, and closing emerging talent gaps required to deploy and sustain AI effectively. Over time, leaders must realign performance expectations, incentives, and career paths to reflect this new model of work.
Effective AI adoption needs more than a road map. AI-first providers must use high-value workflow transformations to fund and shape the enterprise data and AI architecture. This ensures they build reusable capabilities with each deployment and avoid one-off, point solutions that do not strengthen the underlying platform.
Accelerating the Route to AI-First
AI introduces real risks—clinical, operational, and reputational—that must be actively managed through strong governance, validation, and controls. But the greater strategic risk is inaction or incrementalism.
AI-native entrants are already targeting narrow, high-margin value pools with lower-cost, scalable models purpose-built around AI. As these models mature, they will expand selectively, capturing profitable segments while avoiding the full complexity of integrated delivery systems. Incumbents risk being left with a more complex, higher-cost, lower-margin case mix, compounding the rising structural pressures.
To respond, providers must prioritize workflows that deliver near-term impact while building toward a technical and operational long-term target that requires focus, discipline, and a willingness to invest ahead of realized value. They must lead where differentiation and margin defense matter most, while building capabilities to fast-follow elsewhere.
Decisive action now will determine whether AI strengthens the core business—or accelerates its erosion.
