Becker’s caught up with Dennis Chornenky, chief AI adviser at Sacramento, Calif.-based UC Davis Health, for his AI plans and predictions for the year ahead. The conversation has been edited for clarity and brevity.
Question: What do you expect to focus on in AI for 2025?
Dennis Chornenky: I’m focused on three key trends. First, the advent of autonomous AI, particularly AI agents. These are more advanced than generative AI, leveraging large language models to integrate multimodal data inputs — images, sound, labs — and generate outputs like patient histories, summaries, projections, or even presentations sent to specialists. This has implications for workforce displacement, as AI can handle tasks traditionally done by junior professionals. It raises questions about how future workers will gain experience and how we prepare them to work effectively with AI.
Second, integrating AI agents into IT environments presents governance and safety challenges. Traditional governance focuses on predictive AI, but agentic AI introduces new risks, including potential issues with AI-to-AI interactions, similar to drug-to-drug interactions. OpenAI, for instance, is expected to release its first AI agent soon, which will likely kickstart industry adoption.
Finally, the spectrum of AI advancement — from generative AI to artificial general intelligence — is evolving. AGI won’t be a single event but a gradual progression. Future scenarios could include companies entirely run by AI agents, with human oversight limited to strategic roles.
Q: Has AI replaced any healthcare jobs to date?
DC: Not entirely. Instead of replacing roles, AI augments tasks, automating repetitive work and enabling staff to focus on higher-value activities. For example, generative AI tools like ChatGPT are already streamlining processes, freeing up time for more complex tasks. The challenge for organizations is to retrain and upskill workers to collaborate effectively with AI rather than replacing them outright.
Q: Are AI agents currently in use, or are they still in development?
DC: AI agents have existed conceptually for some time, but the integration of large language models is new. These agents go beyond automated workflows, offering more autonomy and decision-making capabilities. For example, an AI agent could handle medical billing, resolve ambiguities, and communicate directly with insurers. This level of autonomy differentiates them from basic workflows. However, governance regulations — defining what they can and cannot do — remain a significant hurdle.
Q: Could AI agents, like those generating reports for specialists, be implemented this year?
DC: It’s possible, but adoption depends on identifying suitable applications within workflows. Early opportunities may lie in coding, summarization, and patient-provider communication. For instance, AI agents could facilitate patient interactions, automate coding, and improve care coordination. Multimodal capabilities will likely drive adoption, but industry discussions on practical applications are still evolving.
