How AI Has And Will Continue To Transform Healthcare

Briar Smith is CTO and cofounder of Attunement.

The Evolution Of AI In Healthcare: Adoption, Hurdles And A Potential Industry Shift

As co-founder and CTO of YC healthcare technology company Attunement, I have experienced firsthand the excitement and challenges of implementing AI within the healthcare system. While AI has already had a significant impact in the automation of tedious tasks, the less-explored but tremendous potential to perform reliable diagnosis and treatment will face significant challenges in adoption and model reliability. While the jury is also still out on to what extent providers will be assisted, or potentially even replaced by AI in the future, more affordable and higher quality care is likely on the horizon—slowly.

The State Of Generative AI In Healthcare

Generative AI has emerged as a promising solution to the U.S.’s broken healthcare system. Its potential applications include everything from assisting clinicians with diagnosis to creating comprehensive treatment plans and even administering therapeutic interventions. Large language models (LLMs) have demonstrated capabilities that can oftentimes parallel human performance in these categories, providing accurate instructions and data.

Under these conditions, they’ve even surpassed medical professionals, such as in the diagnosing of obsessive-compulsive disorder (OCD). At Attunement, our experiments with LLMs for fringe-case diagnoses revealed that while even the best models and clinicians find them challenging, each advancement is rapidly increasing the number of cases that can be accurately diagnosed.

Navigating The Complex Healthcare Landscape

However, the path to implementing these technologies is complex, requiring the navigation of numerous regulatory and logistical hurdles. Products must fall under a valid insurance reimbursement code category and secure approval from insurers while adhering to stringent FDA regulations, which includes demonstrating sufficient clinical efficacy, which can take years for new technology. Proving clinical efficacy is also difficult due to model hallucinations and inaccuracies—lacking reliability compared to a provider with proper medical training.

The current healthcare system presents additional challenges through its misaligned incentive structures between providers and insurers: Providers are paid per service (i.e., for a higher volume of care, not lower). So better outcomes are not incentivized.

Current Applications And Implementation

Due to these constraints, AI’s primary role in healthcare has focused on back-office automation and infrastructure improvements. This includes streamlining intake forms, generating meeting notes, and providing comprehensive summarization services. Companies like DeepScribe have successfully implemented AI for transcription services, while emerging platforms like Thoughtful AI are revolutionizing back-office automation through enhanced payment claims processing and claims management using AI agents. Agents have the ability to take actions and perform work autonomously, similarly to a human, such as by opening applications, copying and pasting, etc.

Language models combined with agents will create artificial general intelligence (AGI), an AI that can do almost anything a human can. Many experts now project that AGI could arrive as soon as 2028, a significant acceleration from previous estimates, due to rapid advancements in the technology. AGI trained on appropriate medical datasets could surpass clinicians’ abilities in diagnosis, treatment planning, and even the administration of treatment.

The Evolution of AI Provider-Assisted Care to…Consumer Personal AI Care?

In the meantime, the healthcare industry is witnessing an emerging trend toward “AI-assisted” care models. This approach involves AI working in conjunction with healthcare providers, offering capabilities such as supporting diagnostic processes, enhancing therapy delivery, providing second opinions on diagnoses and treatment plans and augmenting clinical decision-making

This is likely going to be the medium-term use case until the industry catches up and the technology eliminates remaining hallucinations. This list could expand as the capabilities improve, but the jury is still out on whether over the the long term AI will be an assistive tool for providers or be replaced by consumer-administered care.

The consumer-care theory is one that seems to be looking increasingly likely and favorable. It is care delivered through personalized AI systems tailored to individuals’ characteristics, problems and life circumstances. While this method raises concerns ranging from privacy to ethics, the prospect of having an AI system with the comprehensive understanding of an individual—or at least as much that an AI can have—is likely inevitable, since there is such a need for solutions.

Privacy and security challenges will likely be addressed through the use of on-device models, which will prevent sensitive data from being accessed or commercialized by companies and potential malicious adversaries.

Current Limitations And Challenges

Several significant challenges remain in the widespread adoption of AI in healthcare: The reliability gap persists as models struggle to achieve consistent accuracy in diagnosis and treatment. They remain fundamentally constrained by their input data from providers and often fail to capture subtle nuances in human behavior. Currently, these systems perform best when healthcare providers are trained in AI input symptoms rather than leaving it to the AI to gather this information directly.

Another potential limitation is that clinicians play a crucial role in developing patient rapport and capturing behavioral and emotional data that AI systems currently struggle to assess adequately. It’s unclear whether AI will be able to achieve that or when.

In summary, the adoption of AI in healthcare faces multiple hurdles: Stringent regulatory requirements for new technologies, provider resistance to workflow disruption, complex insurance reimbursement structures for AI-based solutions and risk aversion in the healthcare sector.

Looking Ahead

With estimates as high as $1.9 trillion in annual wasted spending in our current healthcare system, even incremental improvements could yield significant benefits. This potential for positive impact might help overcome the negative connotations often associated with AI in healthcare, particularly to the human interaction side of care and reliability.

The industry’s progress toward value-based care (VBC) insurance models, which prioritize care quality over volume, represents a promising shift. This transition, combined with ongoing improvements in AI reliability and capability, suggests a future where AI plays an increasingly central role in healthcare delivery, despite the regulatory and adoption hurdles. The path forward to success requires careful attention to model reliability and ethics, privacy protections, and the balancing of technological and human elements in healthcare delivery.

Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?