TL; DR: Flashy interfaces and one-off tools may create noise in healthcare AI, but they will not create lasting value. The real opportunity is in platforms that collapse processes, embed into existing systems, and grow into connective infrastructure across clinical, financial, and administrative domains. Furthermore, adoption within healthcare will not only depend on how well AI performs but also on whether the right governance and compliance controls are in place.

AI in Healthcare: Driving Efficiency or Reinventing the Burden?

AI has been widely promoted as the next major shift in healthcare, promising to reduce administrative burden, free up provider time, and streamline inefficiencies. The reality, however, is more complicated. Healthcare is complex not just because of inefficiencies, but because it is shaped by competing incentives, entrenched processes, and regulatory constraints. Prior authorization is not just a broken administrative pathway. It is a cost-control mechanism for payers. EHR fragmentation is not a technical bug. It stems from vendor lock-in and conflicting data ownership. Revenue cycle challenges are not solely about missing automation. They reflect the complexity of navigating government, commercial, and self-insured reimbursement models.

Many AI companies position themselves as the answer to these problems by offering yet another tool or flashy interface. The real challenge, however, is solving inefficiencies without adding new layers of complexity. The best solutions will reshape workflows to be less fragmented and easier to navigate, moving beyond surface-level automation to simplify operations and resolve the pain points that create bottlenecks across the system. This is not only a question of efficiency. It is also about addressing the operational strain that contributes to provider burnout and impacts the quality of care.

What this looks like in practice:

1. AI that fits into existing systems, not outside them

• The initial strongest solutions will integrate directly into core platforms like EHRs and payer portals, keeping work within a single environment instead of across fragmented systems. Once embedded, AI can assist with charting during the visit, auto-populate fields during intake, or initiate prior authorization steps without requiring users to leave the clinical or administrative interface.
• Using AI scribes as an example, players like Abridge, Nabla, and Suki have gained significant initial market share as best-in-class solutions that integrate directly into the EHR. However, ongoing innovation and a pathway to expand into adjacent functions will be essential for continued success, as Epic and Athena’s recent AI scribe launches underscore the risk of potential displacement.
• While there is risk that vendors could restrict integrations over time, solutions that remain best-in-class and indispensable to provider operations are likely to stay sticky, as EHRs are reluctant to cut off tools their customers rely on.

2. AI that eliminates work, not just automates it

• The most impactful AI systems rethink how work gets done by removing unnecessary steps, not just speeding them up. This includes collapsing multi-step processes, bypassing duplicative documentation, and reducing form-filling or manual follow-ups.
• Autonomize, for example, is building an AI coordination engine that not only automates tasks like prior authorization, but also routes information across care teams, payers, and back-office systems. The result is fewer handoffs, clearer priorities, and more streamlined work across the care continuum.
• HealthEx reflects a similar path on the data side. By unifying disparate health datasets into a single interoperable layer, it enables AI to operate across silos and power downstream use cases in clinical care, research, and population health.

3. AI that connects systems, not adds more

• Healthcare does not need a separate AI tool for every use case. It needs intelligence that bridges disconnected systems instead of creating new ones. Many AI solutions may begin with a wedge, addressing a single administrative or clinical task, but the long-term opportunity is to expand across functions and serve as the connective layer between clinical, administrative, and financial systems.
• Abridge exemplifies this trajectory. Starting with ambient documentation and now prior authorization, its platform is evolving to integrate with multiple systems and support end-to-end clinical processes, improving both provider efficiency and downstream data quality.
• Affineon Health is also pursuing this path, automating routine tasks like lab result reviews and prescription renewals in ways that bring disparate workflows together and reduce the need for providers to toggle across systems.

4. AI that works alongside providers, not around them

• The goal of AI in healthcare is not to fully automate care delivery, but to support providers in doing their jobs more efficiently and effectively. The most valuable systems will reduce administrative friction, improve handoffs, and guide decision-making in ways that strengthen, not replace, the provider’s role in patient care.
• For instance, AI can flag abnormal lab values, surface relevant patient history, or suggest evidence-based treatment options, giving providers better information without removing their judgement.
• Atropos Health illustrates this on the clinical decision-making side. Its platform uses real-world data to generate on-demand evidence that informs provider decisions at the point of care, offering guidance in situations where research may be limited without removing the provider’s judgment.

Governance is the Next Bottleneck

As AI becomes more embedded in healthcare, the increasing barrier to broader adoption is no longer technical capability or model accuracy, it’s governance. Findings from the Healthcare AI Adoption Index, published by Bessemer, Bain, and AWS, indicate that while significant capital is flowing into healthcare AI, making it one of the most talked-about and invested-in sectors, deployment at scale is still constrained by the lack of strong governance and accountability frameworks.

Healthcare is governed by a complex and evolving set of regulations that define how patient data can be accessed, shared, and protected. HIPAA is the most well-known, but it is just one part of a broader legal framework that includes federal policies like HITECH, which incentivized the adoption of EHRs and established standards for their meaningful use, and the 21st Century Cures Act, which strengthened interoperability by prohibiting information blocking and requiring open APIs to give patients easier access to their data. These rules place strict requirements on any system that handles protected health information, from how access is granted to how activity is monitored and logged. When AI enters workflows like documentation or billing, even small missteps can carry serious clinical, financial, or legal consequences. Even pilots must clear security and compliance reviews, but these are lighter than the rigorous standards of full deployment. Adoption of these early pilots has been high, yet customers will only expand beyond proof of concept once they have confidence in the solution’s ability to meet strict security, regulatory, and risk requirements.

It is also important to note that AI does not operate in isolation. As models are integrated across workflows, they begin to interact with other systems and, increasingly, with other AI solutions. Whether through agent-based frameworks or distributed APIs, these interactions introduce complexity around coordination, context sharing, and control. When AI makes decisions or triggers downstream actions, those steps must be governed with the same level of rigor as any clinical or financial process. In many cases, however, there is no centralized oversight to ensure those standards are met. Companies like Ferrum Health are beginning to tackle this problem by providing monitoring and governance platforms that track AI performance across clinical operations, ensure compliance with regulatory requirements, and give health systems confidence that these tools can be deployed safely at scale.

Until these frameworks exist, the use of AI to autonomously complete tasks or manage entire workflows will remain limited to narrow, low-risk applications or require human supervision. This is not a technical limitation. It reflects the high trust bar required to operate safely in clinical and financial environments. Widespread adoption will depend as much on governance as it does on performance.

What This Means for the Future of AI in Healthcare

We see three implications that will shape the future of this market:

• Governance will be the foundation for scale. Health systems will only adopt AI at scale from vendors they trust to ensure safety, accountability, and compliance. Governance is not simply a regulatory hurdle; it will separate the pilots from the platforms.
• Platforms will outperform point solutions. While best-of-breed solutions may capitalize on early enthusiasm, the future belongs to companies that collapse steps and connect processes across clinical, administrative, and financial domains. Narrow tools will struggle to grow beyond isolated use cases.
• The winners will give time back to providers. AI that adds friction or distraction will erode adoption. The lasting value will come from solutions that meaningfully expand provider capacity and improve both outcomes and financial performance.

Our initial perspective on AI builds on our broader healthcare thesis. We have long invested in companies that reduce friction, expand provider capacity, and embed technology into care delivery in ways that can scale sustainably.

For example, our portfolio company Sevaro enables neurologists to deliver immediate, high-quality consults by abstracting administrative bottlenecks and automating clinical workflows, while allowing them to produce better clinical outcomes, often in time-sensitive situations. In a similar manner, Tava manages the administrative, clinical, and financial requirements of running a practice so that therapists can be more productive and increase the time they spend providing mental health care that is so desperately needed. Rivet reflects the same philosophy on the revenue cycle side, equipping providers with accurate estimating and forecasting tools that give visibility into expected reimbursement, reduce financial friction, and help practices operate more efficiently.

As AI capabilities proliferate, we see meaningful opportunities for companies like these to embed intelligence into existing operations in ways that strengthen care delivery without adding friction. We are excited to continue backing solutions that align with this vision: ones that work within the system, amplify provider impact, and make healthcare simpler, not harder.