AI is no longer the future of healthcare. The way patients are diagnosed and treated is already changing. Some of the most interesting developments involve systems that sense and respond to human emotions. For example, Cedars-Sinai's Connect platform adapts care based on patient emotions. CompanionMx detects anxiety by interpreting audio and facial cues. And Feel Therapeutics uses emotion-sensing wearables to tailor interventions in real time.
At the same time, clinical tools are evolving. Hospitals are combining large-scale language models (LLMs) with AI note-taking apps like Nabla and Heidi that can listen to, summarize, and respond to the nuances of doctor-patient conversations. Investment in medical scribing technology alone reached approximately $800 million last year.
Transition to AI adaptation
All of this marks a major shift from AI that automates tasks to AI that adapts. Traditional AI has sped up paperwork and data processing. Adaptive AI helps clinicians make better decisions, understand patients better, and respond more appropriately. This shift in breast cancer screening, genomics, and drug discovery is already visible, and high-quality data and continued validation are driving real progress.
When designed responsibly, emotionally sensitive tools can strengthen the connection between clinicians and patients, personalize care, and reduce pressure on overstrained systems. But as adaptive AI becomes more widely available, success will depend less on technical excellence and more on how the system is built. Successful tools will be flexible and adaptable to people, tailored to patient needs, clinician workflows, and the realities of care. Good AI needs to be predictive and context-sensitive, built with full patient diversity in mind.
Of course, even the most empathetic AI cannot erase the imperfections of human systems. For example, recent studies show that medical AI tools and LLM-based assistants routinely downplay women's symptoms and show less empathy for Black and Asian patients than white men. AI does not cleanse real-world bias. It propels them forward and often extends their influence. I've seen this pattern before.
Introduction is important
That's why deployment conditions are as important as the technology. Systems that mimic empathy do not automatically pick up on nuance, context, and risk. Without firm ethical boundaries, so-called emotional intelligence can provide a false sense of security. Clinicians still need to make the final decision to protect patients and maintain trust. AI can be a helpful partner in caregiving, but it cannot take on the burden of human responsibility.
To build trust, you need to strengthen the foundation on which you rely on it. Involving patients, families, and caregivers from the beginning reveals blind spots early and balances compassion with pragmatism. It also becomes clear where automation steps back and human care needs to step in. Our cancer platform, developed in partnership with the Cancer Awareness Trust, demonstrates this in action and shows how empathetic design creates tools that are trustworthy and genuinely useful.
AI is not here to replace humans. We're here to support their expertise and expand their impact. Ideally, we would build machines that handle complexity and pattern recognition, allowing clinicians to focus on what humans do best: exercising judgment, building connections, and providing care. Machines may learn to care, but it is up to us to build ecosystems where that care is trustworthy, fair, and meaningful. It's certainly a challenge, but it's also full of opportunities.
Nikki Splinz is the CEO of ustwo.
The extension deadline for Fast Company's World Changing Ideas Awards is Friday, December 19th at 11:59pm PT. Apply now.
