Future of Healthcare
August 10, 2026
This is the first article in a two-part series on AI in healthcare. Continue to Part 2 to learn about where AI in medicine is headed.
The longevity space is noisy when it comes to AI. The word gets used constantly: as marketing, as shorthand for innovation, as a way to signal modernity. We want to cut through that and be specific about how we actually use AI at Biograph today, why we chose each tool, and what the safeguards look like.
The goal is not AI for its own sake. It is to empower our physicians and our broader team, improve the experience of being a member, and ultimately deliver more personalized care.
AI as a force multiplier for clinical expertise
A comprehensive Biograph visit generates an enormous amount of data: whole-body MRI, CT coronary angiography, a comprehensive panel of blood biomarkers, genetic markers, and a detailed health history. What would traditionally require separate visits to a cardiologist, an internist, a radiologist, and multiple labs, often spread across weeks, happens in a single coordinated session. Reviewing and connecting all of it is exactly the kind of work where the right tools make a real difference.
Dr. Jonathan Fu, Physician at Biograph, compares the shift to when ultrasound transformed central line placements.
"It reminds me of the shift from landmark-based procedures to ultrasound-guided procedures. Physicians trained before ultrasound became routine developed impressive clinical skills using anatomy and landmarks, and many were very good at it. But once ultrasound became widely available, you could directly visualize the needle, the vein, and the surrounding structures. It changed the standard of care. I think AI may create a similar divide: not between good and bad doctors, but between clinicians who learn to use new tools well and those who don’t."
The point is not that technology replaces skill. It is that the best clinicians become even better with the right tools in hand. Ultrasound made already-skilled proceduralists more precise and more efficient. AI does the same for the modern physician, handling the heavy lifting of synthesis so our team can spend its time and attention on the parts of care that genuinely require a human.
Why we don’t apply every AI tool at Biograph
Not every tool that claims clinical value has earned it. Before we deploy a third-party tool at Biograph, we assess its clinical accuracy against our existing practices and against what our team knows about the state of the field. Independent peer-reviewed validation and regulatory clearance, where applicable, are the baseline.
For tools we build or configure internally, the bar is just as high: a thorough test and evaluation process, a controlled beta rollout, and multiple feedback mechanisms that drive continuous improvement over time. We add a capability when we are confident in the benefit, not before.
Where AI shows up in Biograph’s diagnostic journey
We bring specialized, clinically validated third-party AI directly into our diagnostic process. Two clear examples:
Cardiovascular imaging. Standard cardiac screening often misses a critical early signal: coronary artery inflammation. In asymptomatic patients, this inflammation can be invisible on conventional imaging, yet research shows twice as many fatal and non-fatal cardiac events occur in patients without obstructive plaque than plaque assessment alone would predict. Inflammation is often the hidden driver. [1]
Biograph is the first clinic in the United States to study a novel AI-based preventive cardiac marker in asymptomatic patients, through our partnership with Caristo Diagnostics. Clinically eligible Black Tier members can participate in this exclusive research study, which evaluates Caristo's CaRi-Heart® technology: an AI that quantifies coronary inflammation directly from a standard cardiac CT scan, detecting disease activity that precedes visible plaque formation by years. CaRi-Heart is approved for clinical use in Europe and is under U.S. investigational evaluation exclusively at Biograph; CaRi-Plaque™ has received FDA 510(k) clearance.
"Early detection of inflammation is crucial, as it often precedes plaque buildup and provides a critical window for preventing disease."
MRI. A comprehensive whole-body MRI protocol conventionally takes up to two hours, long enough that patient motion becomes a real diagnostic problem. Our 1.5T MRI scanners leverage one of two AI layers that change this.
Siemens Deep Resolve applies deep learning directly to raw k-space data (the frequency-domain signal captured before an image is even reconstructed), achieving up to a 73% boost in acquisition speed according to Siemens during the scan. [2] SubtleMR can be used as a retrospective pass: sharpness enhancement and denoising that restores images to optimal signal-to-noise ratios without additional scanning. A prospective, multicenter, multireader clinical trial found that deep learning reconstruction enables a 60% reduction in scan time while producing images that independent radiologists rated as statistically superior to standard-of-care acquisitions. [3]
The practical result: members spend less time in the scanner. A shorter scan is a more comfortable one, with less time held still and less of the claustrophobia a long study can bring. It also leaves less time for a patient to move, so the images come out cleaner, with fewer repeat acquisitions.
The clinical copilot behind the scenes
Not all of the AI we use touches imaging or diagnosis. Some of the most valuable tools work quietly behind the scenes, giving our clinicians back their most limited resource: time.
We have built an internal clinical copilot that pulls together the data from across a member's record, including MRI, bloodwork, genetics, prior visits, and health history, and synthesizes it into a clear summary. It highlights what matters, surfaces anomalies, and flags trends worth a closer look. Work that could take a physician hours of chart review happens in a fraction of the time, so they can focus on reasoning about the member rather than assembling the picture.
Around that copilot sits a set of supporting tools. When a member sends in outside records, we use AI-enhanced optical character recognition (OCR) extraction to analyze the data, verify it against what we expect (confirming, for example, that a date of birth matches across documents), and integrate it cleanly into our systems. Other tooling helps clinicians draft chart notes, reducing administrative load across the team.
Then there is clinical reasoning. For the genuinely non-algorithmic questions, our physicians use frontier LLMs like Gemini, Claude, and ChatGPT as a thinking partner.
"Medicine is often not purely algorithmic. We rarely have perfect data, and patients bring comorbidities, preferences, and context that make decisions less black and white. I’ll often use AI as a reasoning partner: ‘Give me the strongest argument against this recommendation, and then the strongest argument for it.’ That kind of back-and-forth helps me pressure-test my thinking before advising a patient."
Our physicians also use OpenEvidence, a clinical decision-support platform built specifically for doctors. Unlike a consumer chatbot, OpenEvidence grounds every response in peer-reviewed medical literature, with sourced citations to journals like The New England Journal of Medicine and JAMA. [4] It is the most widely used medical AI among verified U.S. clinicians, and it lets our team pull the current best evidence on a specific clinical question in seconds rather than hours.
It is worth being clear about why this is different from a patient using a chatbot alone. One in three Americans now use AI for health information, yet many still hold legitimate concerns about it: 54% say they want to give explicit permission each time AI is involved in their care. [5] [6] Every frontier model operates under thousands of lines of hidden system-prompt instructions written by technology companies, shaped in part by legal caution and liability management. Those constraints influence answers in ways that are largely invisible to the user. One of the largest and most clinically reputable studies of LLMs in medicine to date – the NOHARM study – found that errors of omission accounted for more than 80% of severe errors when using LLMs for diagnoses and medical advice. [7] This should not be surprising: it is much harder to get into legally unfavorable territory for opting to not provide a diagnosis or advice when you should have, than for giving a false diagnosis or provably incorrect information.
A Biograph physician has no such competing objectives. Their only goal is your health. AI in the hands of a clinician who knows your full history and is accountable to you is a fundamentally different thing from a chatbot used as a substitute for one.
Why physician oversight is non-negotiable
No tool is infallible, and we are candid about that. An AI model can flag something that turns out to be benign, and an imaging algorithm can occasionally misread an artifact. The risk is not only the error itself; it is a member receiving an alarming, uncontextualized finding with no one to help interpret it. That is how well-intentioned technology can still cause harm.
This is exactly why every AI-assisted finding at Biograph runs through a human review chain before it reaches anyone. Our imaging protocols and our radiologists, not an algorithm, own the final interpretation. An AI tool may help surface or quantify something, but a fellowship-trained radiologist reviews it, our clinical team evaluates it in the context of your history and your other results, and a physician decides what it means and how to communicate it. The algorithm surfaces; the physician decides.
Your data stays yours at Biograph
Using AI in healthcare raises real questions about privacy. About 77% of Americans say they are concerned about the privacy of medical information shared with AI tools. [5] Our data governance stance is specific:
Is my data used to train public models? No. We only use AI providers with whom we hold a HIPAA-compliant Business Associate Agreement (BAA) to protect your data. Your personal health information is never used to train external models.
Is my data sold to third parties? No. De-identification is not true anonymization, and we do not rely on it as a workaround. Biograph does not sell member data to brokers, advertisers, or software vendors.
What about the AI tools that do work with member data? Internally all our AI tools, such as our Clinical Copilot, the self-hosted platforms Onyx, and LiteLLM, route every model call through HIPAA-covered infrastructure: Google Cloud and Vertex AI, with whom we hold a BAA. Member data stays inside that secured boundary and is not shared with arbitrary AI vendors.
Who has final clinical authority? A licensed Biograph physician. Always.
Turning data into better decisions
The underlying value of AI at Biograph is personalization. Every member has a unique physiology, history, and set of risk factors. AI lets our physicians engage with that complexity at a level that would not otherwise be possible: synthesizing your entire diagnostic picture, surfacing what is relevant, and building an intervention plan specific to you rather than one based on population averages.
That picture may include everything from whole-body MRI and cardiovascular risk assessments to blood biomarkers, genetic risk factors, body composition analysis, and fitness testing, all integrated into a single longitudinal view of health.
Better diagnostics, faster imaging, more informed physicians, and a more efficient team all point to the same thing: more time and attention on what actually matters – your individual health trajectory.
Technological innovations and AI have given each of us unprecedented access to thousands of health data points, but more health data on its own does not make anyone healthier. Left uncontextualized, it can even create anxiety, cause confusion, and lead to massive amounts of misinformation about what it means to be “healthy.” Data becomes valuable only when it is interpreted and transformed into a clear, personalized, and actionable plan. That is the role AI plays at Biograph. It sharpens our tools, allowing our physicians and care team to focus their expertise on the thing that matters most to us: your health.
Commonly asked questions about AI at Biograph
Does AI make medical decisions at Biograph?
No. AI helps analyze information, surface findings, and streamline workflows, but all clinical decisions are made by licensed Biograph physicians.
Does Biograph use AI to interpret imaging scans?
Biograph incorporates clinically validated AI tools into certain imaging workflows, including cardiovascular imaging and MRI reconstruction. However, all imaging findings are reviewed and interpreted by qualified radiologists and physicians.
How does AI help physicians at Biograph?
AI helps physicians synthesize complex information from imaging, bloodwork, genetics, health history, and prior visits. This reduces administrative burden and allows clinicians to spend more time focused on patient care and clinical decision-making.
Does AI replace physician expertise?
No. AI is designed to support physicians, not replace them. Biograph uses AI as a tool to enhance clinical efficiency, improve data synthesis, and support personalized care.
What AI technologies does Biograph use today?
Biograph uses AI across several areas, including MRI image enhancement, cardiovascular imaging analysis, clinical documentation, medical record processing, and physician decision support.
Can AI identify health risks before symptoms appear?
Some AI-enabled diagnostic tools may help detect patterns associated with disease risk earlier than traditional approaches alone. Findings are always reviewed in the context of a member’s complete clinical picture by a physician.
How does Biograph evaluate new AI tools?
Biograph evaluates AI technologies based on clinical evidence, validation studies, regulatory status where applicable, internal testing, and physician review before implementation.
Jack Larch is Biograph’s Lead AI Engineer, leading the development of AI systems and internal tools that help both clinicians and internal teams work more efficiently, synthesize complex health data, and deliver more personalized care.
Clinical references
Chan et al., "Inflammatory risk and cardiovascular events in patients without obstructive coronary artery disease: the ORFAN multicentre, longitudinal cohort study." The Lancet, 2024. https://doi.org/10.1016/S0140-6736(24)00596-8
Siemens Healthineers. Deep Resolve. Accessed June 2026. https://www.siemens-healthineers.com/en-us/magnetic-resonance-imaging/technologies-and-innovations/deep-resolve
Deep learning enables 60% accelerated volumetric brain MRI while preserving quantitative performance: a prospective, multicenter, multireader trial. AJNR Am J Neuroradiol. 2021;42(12):2209-2215. doi:10.3174/AJNR.A7344
OpenEvidence. About OpenEvidence. Accessed June 2026. https://www.openevidence.com/about
KFF. 1 in 3 adults are turning to AI chatbots for health information, equaling the share who use social media for health. Published May 14, 2025. Accessed June 2026. https://www.kff.org/health-information-trust/poll-1-in-3-adults-are-turning-to-ai-chatbots-for-health-information-equaling-the-share-who-use-social-media-for-health/
Verasight. AI in healthcare: public perceptions and attitudes. Presented at the American Public Health Association Annual Meeting; 2025. Accessed June 2026. https://nativereach.com/ai-in-healthcare/
Wu D, Haredasht FN, Maharaj SK, et al. First, do NOHARM: Towards clinically safe large language models. arXiv. Published December 1, 2025. doi:10.48550/arXiv.2512.01241
