By Allison Proffitt
November 17 2026 | SAN FRANCISCO — Walter Greenleaf had planned to deliver his keynote at the Neurotech Leaders Forum last week in San Francisco in person. Instead, the Stanford neuroscientist appeared by video from Hong Kong, where he had landed about an hour earlier. The message he brought to the forum’s 25th-anniversary gathering was one he has been refining since what he called “the Apple II days”: digital health is no longer a bolt-on to neurotechnology. It is becoming the layer through which neurotech is delivered, managed, and paid for.
Greenleaf, a research scientist with Stanford’s Virtual Human Interaction Lab and a 35-year veteran of applying virtual reality and AI to clinical products, opened with what he stressed was a premise rather than a finding. As technology companies compete to build the conversational AI agents that will mediate how people interact with each other, with devices, and with the built environment, he predicted, health care will become one of the services those agents quietly provide in the background.
“Healthcare infrastructure will be part of infrastructure in general,” he said, because the same systems that read a user’s requests will also read voice tone, facial expression, and other biosignals.
In fact, Greenleaf foresees significant predictive and diagnostic applications will soon be delivered through digital health tools and ecosystems. For instance, while acknowledging that reliably predicting mental health crises such as suicidality will take time, Greenleaf singled out work using wearable sensors to anticipate multiple sclerosis flare-ups as an example of where early warning could cut costs dramatically, especially for patients far from major medical centers.
While he didn’t have a short term reimbursement model in mind, in the long run, Greenleaf expects that companies such as Apple, Google, Samsung, and Amazon, competing for consumers’ attention, will offer health-crisis prediction as a subscription service, or that insurers will buy it. The reason someone chooses one company’s smart glasses over another’s, he suggested, may ultimately be how well they keep that person healthy.
A Confluence, Driven by Demographics
Greenleaf framed the opportunity against the pressure of an aging population. The U.S. population has doubled since the 1970s while family sizes have shrunk, he noted, leaving fewer informal caregivers to support older adults facing higher rates of Alzheimer’s, Parkinson’s, cancer, and heart, lung and kidney disease. Quoting William Gibson, he said solutions already exist but are “not very evenly distributed”; the challenge is scaling them profitably and equitably.
He identified five trends he expects to shape the next wave of digital health:
- Ambient AI that turns clinical documentation into a background process rather than a burden.
- Agentic workflows that move from recommending tasks to executing multi-step administrative and clinical work.
- Immersive technologies and smart glasses, particularly devices with inward-facing cameras that track the face and eye gaze, which can both record and intervene.
- Continuous monitoring through patches, home sensors, and wearables, generating the data to measure interventions and sort patients into subtypes.
- Digital twins, borrowed from industry, enabling “predictive maintenance of our bodies and our minds” for individuals and, in aggregate, for families, communities, and nations.
Together, he said, these let care shift from episodic to continuous, closing a loop of sensing, interpreting, intervening, and evaluating.
Devices That Ship With Software
For the neurotech industry specifically, Greenleaf predicted that devices with clinical benefit will increasingly be issued alongside app-based systems designed to boost their efficacy, producing combination products that can be managed and updated in the field. The most important function of that software layer, in his view, is adherence. Borrowing from the gaming and entertainment industries, developers are learning to design apps that draw on the brain’s reward systems to keep patients engaged.
That software must not add to clinicians’ workloads, he cautioned. The goal is a background system that watches for problems and flags only actionable moments between visits.
Greenleaf also argued that a smartphone already qualifies as a medical wearable, and that ordinary cameras can yield clinically meaningful signals. Eye-movement data captured by a phone or webcam, he said, can estimate an individual’s probability of developing Parkinson’s disease and support medication management. He recounted a video call with the CEO of a company holding a large eye-gaze dataset who, at the end of the conversation, showed Greenleaf a histogram of his own attention, fatigue, and distractibility mapped against the discussion. The CEO did not share the equivalent data on Greenleaf, which, Greenleaf noted, “would have been a bit off-putting.”
He pointed to mental health as the area most in need of validated tools, citing the work of Stanford colleague Leanne Williams, who leads the Stanford Center for Precision Mental Health and uses imaging and sensor data to identify biotypes of depression and match patients to interventions. He also highlighted a less technical idea: using digital tools to wrap a “layer of narrative storytelling” around a patient’s care, which he distinguished from gamification, to improve retention, motivation, and engagement.
Citing investment as a signal of momentum, Greenleaf said digital health ventures received more than $14 billion across 482 deals in 2025, and that the field is moving “from hype to operational integration” and into what he called an outcomes era.
The Cost Question Cuts Both Ways
Greenleaf was blunt about the industry’s shortcomings. Many early-stage companies, he said, fail to design for the realities of busy clinics with limited time and budgets, down to basic questions like “who’s going to charge the batteries?” Products that show measurable benefit in controlled settings often “fall apart” when scaled across an enterprise.
He warned that digital neurotech can raise spending as easily as lower it. Earlier detection, hospital-at-home models, telemedicine, and broader behavioral health access can cut costs. But more measurement also produces false positives, incidental findings, alert fatigue, and defensive referrals, and shifts screening and treatment costs forward before savings arrive. Implementation itself carries costs in cybersecurity, workflow redesign and training.
Among the gating factors to adoption, he listed:
- Local validation and continuous monitoring of models for accuracy, bias, and drift across populations and settings.
- Thin evidence: too few randomized controlled trials, small sample sizes and understudied populations, and too little work connecting measurable signals to actual patient benefit.
- A perception problem, as adherence features lead some to dismiss digital therapeutics as games.
- A “huge interoperability problem” across a proliferating array of sensors, analytics, and platforms.
- Protection of neural data from the outset, through consent, limits on secondary use and restricted data retention.
Greenleaf urged attendees to read The Battle for Your Brain by Duke law professor Nita Farahany, which examines the right to cognitive liberty in the age of neurotechnology. He also cautioned against cluttering patients’ lives with too many devices, interfaces, and settings, calling that a design consideration the industry must address now.
Asked whether digital health has more to gain from partnering with pharma or with medtech, Greenleaf pointed to pharma activity such as Takeda’s moves with sensing-and-analytics company Beacon and Alto Neuroscience’s use of digital tools to identify patient subtypes. But he said device companies are “moving much faster” in embracing digital technology, because their roots are in electronics while pharma’s are molecular.