The Jevons Paradox In Medicine: Why Cheaper Diagnostics Mean More Care
- David Priede, MIS, PhD
- 5 hours ago
- 5 min read
Hospitals are moving away from software subscriptions and paying directly for automated clinical work.
Doctors invest many hours daily in billing and data entry. Transitioning from software subscriptions to payments for AI-driven tasks can ease this bottleneck, allowing a renewed focus on the essential patient-doctor relationship.
Takeaways
Hospitals will buy agentic work, not software seats.
Doctors will return to relational patient care.
Cheaper diagnostics lead to more preventative screenings.
Single physicians can run lean, tech-forward clinics.
Graph lookups reduce computing costs and protect data.
The Token Economy: How Agentic Work Changes Healthcare
The software industry is changing how it sells products. Instead of selling a subscription "seat" for a human to use a program, companies are selling "tokens" to pay an artificial intelligence agent to do the work. This shift from access to action carries cascading effects for medicine. Hospitals are beginning to buy agentic work to circumvent the clinical bottleneck that has constrained patient care for decades.
The Science: From Administrative Diagnostics to Relational Care
Over a decade ago, federal mandates forced hospitals to transition from paper charts to Electronic Health Records. The goal was to improve care, but the actual result was an administrative crisis. Physicians became highly paid data entry clerks.
Doctors now spend a large portion of their day on non-clinical tasks such as data entry, insurance coding, and basic compliance. This administrative burden removes them from 3the patient.
We can see the contrast clearly when comparing the old system to the new architecture.
The historical baseline: A hospital buys software subscriptions. The software requires manual human input. Doctors spend hours typing into the system, clicking through complex drop-down menus to satisfy billing requirements.
The token model: A hospital buys agentic work. Artificial intelligence (AI) systems listen to the patient encounter, generate the charts, and handle the billing codes. They operate as background digital assistants. The hospital pays a fraction of a cent per computational token for the machine to complete the task. The culmination of this process is a finished medical chart, ready for the physician to review.
This shift returns medicine to the relational sector. The relational sector relies on human connection. When a patient is sick, they seek human empathy, reassurance, and physical touch. The machines handle the digital bricks of medical charting.
The physicians shift their focus back to the human-mediated experience, offering complex diagnoses and personalized treatment plans that machines cannot replicate.
An Expert's Perspective: The Economics of Diagnostics and Care
When a task becomes cheaper, people often do more of it. Economists call this the Jevons Paradox, named after a nineteenth-century observation that more efficient steam engines led to higher coal consumption. If artificial intelligence makes analyzing a magnetic resonance imaging (MRI) scan or a genomic sequence ninety percent cheaper, hospitals will not fire their radiologists. They will run more tests.
The science is sound. This democratizes diagnosis. We can look at the objective evidence of how this changes clinical settings.
Proactive Screenings: Lower costs allow doctors to order deeper analyses that were previously too expensive or too bespoke to justify. A genomic test once reserved for severe cases can become a standard baseline check. This leads to earlier detection of pathology and a higher volume of preventative care.
Medical Startups of One: We are seeing an unprecedented drop in computing costs. We will soon see a rise in highly specialized, lean medical practices. A single nurse practitioner, supported by a network of automated agents, can manage a patient load that once required a full administrative staff.
Accessible Longevity Tech: Longevity medicine focuses on tracking biomarkers to extend human health. Small, tech-forward firms can use automated systems to track complex aging protocols for patients. By using artificial intelligence tokens to coordinate care, they can offer this highly customized service at a fraction of the cost of traditional concierge medicine.
Efficient Data Retrieval: Companies are learning to use graph lookups instead of vector dumps to retrieve data. Instead of a costly, token-hungry search through thousands of pages of unstructured medical notes, a knowledge graph maps out the relationships between data points.
Improved Patient Safety: By using graph lookups, an automated agent can instantly find relevant drug interactions or patient history. This addresses fragmented medical records and ensures the doctor has accurate information before prescribing medication.
To understand the computing side, we look at how technology companies build their internal systems. Organizations map out work using specialized data layers.
For example, Atlassian uses a system called a Teamwork Graph to map how people, projects, and documents connect across their platforms . When we apply this logic to healthcare, a medical knowledge graph maps the connections between a patient, their lab results, and their historical diagnoses.
Because the relationships are clearly mapped, the artificial intelligence agent does not have to guess. It pulls the exact data required, spending fewer tokens and returning an accurate, factual summary. This prevents the system from generating false information, a critical requirement for patient safety.
The Road Ahead
The goal is to build up the people working in clinical settings, not replace them. Automated agents will remove the drudgery of the job. This allows medical professionals to operate at the top of their license, focusing on the high-level decision-making that requires human judgment.
However, we must balance this enthusiasm with scientific caution. A breakthrough in software architecture is a game-changer, but we face real-world constraints. The United States Food and Drug Administration (FDA) strictly regulates medical software. The agency is currently building frameworks for how to monitor algorithms that learn and adapt over time, requiring pre-determined change control plans.
We must establish clear liability rules for when an automated system makes a billing or diagnostic error. Who is responsible when an autonomous agent makes a mistake? Data privacy remains a massive hurdle, as hospitals must protect patient records while feeding them into large computational models. Physician training must also change, teaching doctors how to oversee these agents rather than performing the manual work themselves.
The long-term human impact depends on how we build the infrastructure. Treating machines as assistants rather than authorities keeps patient care grounded in human empathy. That is how the clinical system will change.
FAQs
What is the difference between seats and tokens?
A seat is a subscription for a human to use software. A token is a payment for a machine to do the work.
How does artificial intelligence help doctors?
It handles administrative tasks like typing charts and assigning billing codes, giving doctors more time for patient care.
What is the Jevons Paradox in healthcare?
When medical analysis becomes cheaper and faster, hospitals will run more tests rather than firing their staff.
What is a medical startup of one?
It is a small, specialized clinic run by a single doctor who relies on automated agents instead of administrative staff.
Why are knowledge graphs better than vector dumps?
Knowledge graphs map the exact relationships between medical data points, making data retrieval cheaper, faster, and more accurate.
