"AI-native EMR" is often discussed as a single category, but the role such a system should play changes substantially depending on whether it is deployed in a clinic or a hospital. A small outpatient clinic and a several-hundred-bed acute care hospital face different constraints, and the work that humans should be doing differs as well. This article organizes these two directions along the axis of "clinics aim for unstaffed operation, hospitals aim for creating room," and examines why this matters for how Japan works overall.
Disclaimer: This article provides general information. Statistics reflect figures as published and are subject to revision. Always verify against current primary sources (the Statistics Bureau of Japan, the Ministry of Health, Labour and Welfare, and so on).
For the definition of an AI-native EMR itself, see What Is an AI-Native Electronic Medical Record?.
First, the Scale of the Healthcare Workforce
Before the argument, the numbers. How much it matters to change the way healthcare works depends on this scale.
According to the 2025 annual average of the Labour Force Survey published by the Statistics Bureau of Japan, Japan had 68.28 million employed persons. Of these, 9.47 million worked in the industry category "medical, health care and welfare"—roughly 13.9%, or about one in seven workers.
What stands out is the growth. In the 2025 average, "medical, health care and welfare" was the industry that added the most workers year over year: +250,000. Over the same period, manufacturing fell by 130,000 and wholesale and retail trade by 160,000.
| Industry | Employed persons (2025 avg.) | Year-over-year |
|---|---|---|
| Manufacturing | 10.33 million | −130,000 |
| Wholesale and retail trade | 10.29 million | −160,000 |
| Medical, health care and welfare | 9.47 million | +250,000 |
| Services (not elsewhere classified) | 4.82 million | +160,000 |
| Information and communications | 3.02 million | +100,000 |
Be Careful with the "Ten Percent" Framing
The scale of the healthcare workforce is sometimes described as "roughly ten percent of all workers," but that figure changes depending on what you count.
The 13.9% above is the industry category "medical, health care and welfare," which includes long-term care and welfare. Indeed, most of the recent growth comes from the care and welfare side. Isolating the medical field alone yields a smaller share. In the detailed tabulation of the 2020 Population Census, of the 7.63 million workers in "medical, health care and welfare," medical services—hospitals, general clinics, and dental clinics—and elderly welfare and long-term care services each account for a substantial portion; narrowing to the medical field puts the share in the neighborhood of 6%.
Headcounts for the major clinical professions are as follows.
| Profession | Number | Source / point in time |
|---|---|---|
| Nursing personnel (nurses, assistant nurses, public health nurses, midwives) | approx. 1.698 million | End of 2024, employed |
| Physicians (working at medical facilities) | approx. 331,000 | 2024 |
| Pharmacists (registered) | approx. 329,000 | As of December 31, 2024 |
In short, "healthcare workers make up ten percent of the workforce" is an understatement if care and welfare are included (the actual figure is about 14%) and an overstatement if limited to the medical field. Under either definition, this is a segment the Japanese labor market cannot ignore.
The Share Will Grow Further by 2040
Ministry of Health, Labour and Welfare projections indicate that the medical and welfare field will require roughly 10.70 million workers by 2040 (about 18–20% of all employed persons), while only 9.74 million can realistically be secured—a shortfall of about 960,000. From 8.23 million (12.5%) in 2018, the ratio continues to climb steadily, bringing into view an era in which one in five workers is connected to medical and welfare services.
In a society with a shrinking workforce, this field alone must keep adding people. That is precisely where the necessity arises to reduce the work itself through technology.
For Clinics: An EMR That Runs with as Little Staffing as Possible
What clinics face—particularly small outpatient practices in rural areas—is a problem that precedes efficiency. It is that they cannot hire at all.
A staff of one or two administrative personnel is not unusual. Reception, phone handling, intake questionnaires, payment, and claim processing are all run by a handful of people, so a single resignation or illness destabilizes the practice. A director reviewing claims between consultations is an everyday sight.
In this situation, what an EMR is asked to do is not "make staff work faster." It is to create a state in which care can continue even without staff.
Concretely, the goal is for the following flow to hold without human intervention.
- AI handles patient bookings and phone calls
- Pre-visit questionnaires arrive as structured data and connect to the chart
- The consultation conversation converts directly into documentation, orders, and paperwork
- Billing checks and claim generation run automatically, heading off rejections and reductions in advance
- Payment and claim submission connect end to end
This is not a story about convenience. It is a question of whether rural medicine can avoid shutting down. If a practice can function with one physician and minimal staff, that community keeps its healthcare. Unstaffed operation is not a cost-cutting device; it is directly tied to the sustainability of regional care.
For the broader picture of AI adoption on the clinic side, see How to Advance Clinic DX: Where to Start, Priorities, and Adoption Steps.
For Hospitals: Creating Room
Bringing the same thinking into a hospital misses the mark. A hospital is not a place aiming to be unstaffed.
Hospitals bring together physicians, nurses, pharmacists, laboratory and radiology technologists, rehabilitation staff, medical clerks, clinical assistants, and administrative departments, and delivering care as a team is itself the value. Staffing standards impose regulatory constraints as well, so reducing headcount cannot be the objective.
What, then, should an AI-native EMR do in a hospital? Cut the work that is not facing the patient.
A great deal of on-the-ground time is consumed by activity that is not care itself: transcribing records, handovers between departments, drafting referral letters and certificates, discharge summaries, various filings and reports, preparing materials for meetings, duplicate entry across systems. All of it is necessary, but it is not necessarily work that only that profession can do.
Cutting it yields two things.
First, focus on the patient in front of you. Attention previously consumed by documentation and administration can return to the patient. Quality of care ultimately comes down to how fully one can engage with the person.
Second, room. This tends to drop out of efficiency discussions, but it is essential. If the freed time is spent packing in more patients, nothing on the ground changes. Instead, the point is to create time that can be directed toward creative work—examining cases in depth, developing junior staff, redesigning how work is organized, pursuing clinical research, envisioning new forms of care.
Physician work-style reform has advanced through the "tighten the ceiling" approach of overtime caps, but capping hours while the total volume of work stays constant only squeezes the front line. It cannot hold without a means of reducing the work itself. This point is also addressed in Physician Work-Style Reform and Operational Efficiency with AI.
The Two Directions Side by Side
| Dimension | For clinics | For hospitals |
|---|---|---|
| Core problem | Cannot hire in the first place | Excess work per profession |
| Target state | Runs with as little staffing as possible | Able to focus on patients, with room to spare |
| Scope of AI | Substitutes for the whole reception-to-billing workflow | Trims peripheral work per profession and supports the team |
| Reducing headcount | Close to the goal (a means of survival) | Not the goal (staffing standards and team care are premises) |
| Measure of success | Can care continue with minimum staff? | Did patient-facing and creative time increase? |
| Social meaning | Survival of regional healthcare | A shift in the quality of work |
Starting from the same product philosophy, the optimal answers diverge this far. The important point is that neither "scale up the clinic version to get the hospital version" nor "strip down the hospital version to get the clinic version" holds.
Why This Amounts to Changing How Japan Works
Put these two together, and AI adoption in healthcare is more than intra-industry efficiency.
As shown above, medical and welfare services account for about one in seven workers, approaching one in five by 2040. When the way a segment of that size works changes, that is a story about how Japan works.
Moreover, healthcare is not only labor-intensive but a domain where "how many people to staff" is prescribed by regulation through staffing and facility standards. Conversely, when both regulation and operational practice move, change spreads across the field rather than remaining confined to individual workplaces.
Workplaces built on the premise of long hours become free of documentation and administration, and can spend time exercising expertise and doing creative work. If healthcare can achieve that, the same structure becomes visible in other industries facing similar conditions. The significance of pursuing AI in clinical settings lies in this range.
The Shared Foundation: Not Bolted On
The target states differ between clinics and hospitals, but the precondition for reaching either is shared: AI added as an afterthought reaches neither level.
Adding a voice-input button or a summarization button to an existing EMR yields improvements in input and output alone. Neither "it runs without people" nor "peripheral work disappears" is achievable unless AI is continuously present within the workflow, with data structures and UI designed around that premise.
- In clinics, unless reception through billing connects as a single flow, human intervention will always be required somewhere
- In hospitals, only by understanding work whose context differs by profession can the right work be cut
Where this difference originates is covered in detail in What Is an AI-Native Electronic Medical Record?.
Conclusion
- Workers in "medical, health care and welfare" number 9.47 million, or about 13.9% of all employed persons (2025 average). The growth is the largest of any industry
- The "ten percent" framing is definition-dependent: about 14% including care and welfare, around 6% narrowed to the medical field
- By 2040, the field will require roughly 10.70 million workers (18–20%), with a projected shortfall of about 960,000
- For clinics, the goal is operation with as little staffing as possible—a condition for sustaining rural healthcare
- For hospitals, the goal is time to focus on patients and room for creative work
- Changing how a segment of this size works is equivalent to changing how Japan works
Pottech is developing AI-native EMRs addressed to each of these two directions. For inquiries or a demo, please contact us.
References
- Statistics Bureau of Japan, "Labour Force Survey (Basic Tabulation), 2025 Annual Average Results: Summary"
- Ministry of Health, Labour and Welfare, "White Paper on Health, Labour and Welfare 2020" / simulation of employment in the medical and welfare field
- Ministry of Health, Labour and Welfare, "Report on Public Health Administration and Services 2024 (employed health professionals)"
- Ministry of Health, Labour and Welfare, "Survey of Physicians, Dentists and Pharmacists 2024"
- Statistics Bureau of Japan, "2020 Population Census, Detailed Tabulation"