OVERVIEW
AI drafts documents from your medical records
Four stages, divided among robots, AI, and people
- 1RPACollect recordsThe robot pulls records from the electronic health record (EHR)
- 2AIDraft the documentAI reads the records and drafts the document field by field
- 3Physician or nurseReview and editThey check and correct the content, then finalize the document
- 4RPAExport and sendThe robot turns it into a PDF, then saves, sends, and prints it
Before anything leaves your facility, a person reviews and edits it.
SampleFrom record to finished document
A) Dementia slowly progressing. BP stable. Low back pain not worse
Home care guidance report
Client
Patient A (fictional)
Visit date
Tue, Sep 8, 2026
Client’s condition and clinical course
Dementia is progressing slowly. Blood pressure is stable, and low back pain has not worsened. If the nighttime behavior continues, we will consider reviewing her oral medications.
PDF exportedSent via care platform
RPAFields the robot fills inAIRecord excerpts and draftsPersonReview and edit
The medical institutions, care agencies, and patients in the diagrams are fictional, for illustration only.
Home-care clinic
-
Home care guidance reportDeployed
After each visit, drafts the physician’s report to the patient’s care manager, then exports it to PDF and sends it.
-
Home-visit nursing orderIn progress
For the physician’s order that authorizes home-visit nursing, drafts the patient’s condition from a month of medical records and lists the items the physician needs to check.
-
Home care planIn progress
Drafts the “Home visit plan” field from a month of medical records.
-
Extracting and compiling outpatient data submission add-on itemsDeployed
Picks out eligible items from the medical records and builds the submission list.
-
Extracting home care data submission add-on itemsExample
Organizes eligible items from home care records in preparation for submission.
Home-visit nursing station
-
Home-visit nursing reportDeployed
From a month of nursing records, drafts the “Clinical course” section of the monthly report to the attending physician, then produces the report as a PDF.
-
Chart entry formattingDeployed
Tidies voice-dictated home-visit nursing records into readable entries and adds them to the medical record.
-
Incident reportsExample
Organizes the facts and sequence of events to report, and helps draft the report.
Hospital billing and management
-
Extracting and compiling outpatient data submission add-on itemsDeployed
Picks out eligible items from the medical records and builds the submission list.
-
Length of stay by bed typeExample
Calculates length of stay for each bed type and compiles a list for review.
-
Claims checksExample
Cross-checks against the items you set, helping staff review insurance claims (receipts).
-
Payment reconciliationExample
Organizes payment information to help staff reconcile payments.
-
Analyzing management KPIs and posting them to SlackExample
Organizes management KPIs and helps post the results to your chat tool.
Hospital wards and community liaison
-
Discharge summariesExample
Organizes and summarizes clinical records into a draft.
-
Discharge support plansExample
Organizes the information needed and helps prepare a draft that follows your facility’s form.
-
Alerts for inpatient prescriptions about to run outExample
Sends alerts that help staff catch prescriptions about to run out. Healthcare professionals make the prescribing decisions.
-
Alerts for routine test ordersExample
Organizes information that needs checking and alerts the staff member in charge.
-
Unreviewed findings listExample
Compiles unreviewed findings into a list staff can work through.
Home-care clinic: Clinics that provide home medical care can use it for the documents written after each visit and for extracting eligible items for the data submission add-ons (extra fee items in Japan’s medical fee schedule). Support for home-visit nursing orders and home care plans is in progress.
Home-visit nursing station: At home-visit nursing stations (agencies whose nurses care for patients at home), we mainly help with month-end reports and with tidying day-to-day records.
Hospital billing and management: In billing and management departments, we help with extracting and compiling outpatient data submission add-on items, preparing lists for review, and more.
Hospital wards and community liaison: Wards and community liaison offices (which coordinate referrals, admissions, and discharges with outside providers) can use it to draft discharge documents and to send alerts about information that needs checking.
For tasks marked “Example,” we ask each facility about its procedures and design the process to match. See all 19 supported tasks
How this diagram works
Choosing a facility or department shows only the tasks that apply. All the task text is in the page itself, so even where scripts don’t run, making a selection shows the same content.
radio + :has() (switching without scripts) / View Transitions (swapping the lists) / @container (switching the tag and list layout)
In older browsers, the four lists are stacked vertically, each labeled with its facility or department.
*“Deployed” tasks have been in operation at client sites. “In progress” tasks are being prepared for operation at client sites. “Example” tasks show the kinds of work we can handle. Status as of September 2026.
See all 19 supported tasks in one listDeployed4 tasksIn progress2 tasksExample13 tasks
“Deployed” tasks have been in operation at client sites. “In progress” tasks are being prepared for operation at client sites. “Example” tasks show the kinds of work we can handle.
Preparing documents 7 tasks
Organizes medical and work records and drafts documents from them.
-
Home care guidance reportDeployed
Drafts the report from the day’s home visit notes. After the physician reviews and edits it, RPA exports it to PDF, uploads it to the regional care-coordination platform, and faxes it.
- InputThe day’s home visit notes
- AIDraft for each field
- PhysicianReview and edit
- RPAPDF, care platform, fax
-
Home-visit nursing reportDeployed
Drafts the “Clinical course” section from the month’s nursing records. It also helps enter the draft into the report, export the PDF, and update the list of patients due for reports.
- InputThe month’s nursing records
- AIDraft of “Clinical course”
- NurseReview and edit
- RPAReport, PDF, patient list
-
Home-visit nursing orderIn progress
From a month of medical records, drafts “Condition and treatment status” and other fields of the physician’s order for home-visit nursing. It then updates the date and medication fields, and exports and prints the PDF.
- InputA month of medical records
- AIDraft of “Condition and treatment status” and more
- PhysicianReview and edit
- RPAUpdate date and medication fields; PDF and print
-
Home care planIn progress
Drafts the “Home visit plan” field from a month of medical records. RPA enters it into the form, saves it as a PDF, and prints it.
- InputA month of medical records
- AIDraft of “Home visit plan”
- PhysicianReview and edit
- RPAFill in the form; save PDF and print
-
Discharge support plansExample
Organizes the information the plan needs from the records and helps prepare a draft that follows your facility’s form.
-
Discharge summariesExample
Organizes and summarizes clinical records into a draft discharge summary. Healthcare professionals review, edit, and finalize it.
-
Incident reportsExample
Organizes the facts and sequence of events to report, and helps draft the report. The staff member in charge reviews it before submitting it.
Organizing records and tallying data 4 tasks
Pulls out the information needed, tidies records, and helps prepare data for submission and review.
-
Chart entry formatting (home-visit nursing records, etc.)Deployed
Tidies chart entries so they read clearly. In home-visit nursing, it is used to organize voice-dictated records and enter them into the medical record, vital sign readings included.
- InputVoice-dictated nursing records
- AIFormat text and organize readings
- NurseReview and edit
- RPAUpdate the medical record
-
Extracting and compiling outpatient data submission add-on itemsDeployed
Extracts eligible items from the medical records and builds the submission list. The staff member in charge reviews it before submitting it.
- InputMedical records
- AIExtract eligible items
- Staff in chargeReview
- RPASubmission list
-
Extracting home care data submission add-on itemsExample
Organizes eligible items from home care records and helps prepare data for submission.
-
Length of stay by bed typeExample
Calculates length of stay for each bed type and compiles a list for review.
Flagging items and sending alerts 4 tasks
Organizes information that needs checking, based on criteria your facility sets, and alerts the staff member in charge.
-
Alerts for routine test ordersExample
Organizes the routine test order information that needs checking and sends alerts. We design the criteria and how alerts are sent around your facility’s workflow.
-
Alerts for inpatient prescriptions about to run outExample
Sends alerts about inpatient prescriptions to help staff spot those about to run out. Healthcare professionals decide on prescriptions and make any changes.
-
Identifying at-risk patientsExample
Organizes the patient information to be reviewed, based on criteria your facility sets. Healthcare professionals review the extracted information and decide how to respond.
-
Unreviewed findings listExample
Compiles unreviewed findings into a list and outputs it in a format staff can use to review them.
Billing, payments, and management 4 tasks
Also useful for billing checks, retrieving documents, and organizing and sharing management KPIs.
-
Claims checksExample
Cross-checks information against the items you set, helping staff review insurance claims.
-
Downloading and printing invoicesExample
Automates retrieving invoices from business systems, saving them, and printing them.
-
Payment reconciliationExample
Organizes payment information to help staff reconcile payments. We design the process around the systems you use and your facility’s procedures.
-
Analyzing management KPIs and posting them to SlackExample
Organizes and analyzes management KPIs and helps post the results to Slack. We set which KPIs to tally and what to share based on your facility’s workflow.
Tasks done entirely through on-screen operations, such as downloading and printing invoices, can also be automated with RPA alone. See RPA implementation and support
These are just some of the tasks we can handle. Staff review and edit the documents and lists AI produces before using them. Physicians make the final decisions on care and prescriptions, and the staff in charge make the final decisions on billing.
Talk to us about other documents and everyday computer work, too. We tune the RPA processing and the AI’s instructions to your systems and procedures. After rollout, you can extend it to more tasks whenever you like.
DOCUMENTS
From medical record to delivered document
Choose a document, and the robot, AI, and people work through their steps on screen, one after another. You can pause partway or go one step at a time.
To watch again, turn on Autoplay.
Support for home-visit nursing orders is currently in progress.
Home care guidance report (physician → care manager) – Deployed
S) Daughter: “She gets up at night and rummages through the chest of drawers 2–3 times a week1”
O) BP 132/76, P 72, SpO2 97%. Walks indoors holding onto furniture
A) Dementia slowly progressing. BP stable. Low back pian not worse1Typo
P) Fall risk. Discuss raising day svc 2→3x/wk with CM2Abbreviation
ACP) Patient: “I want to stay at home as long as I can.” Daughter agrees. Wants ambulance transport if condition suddenly worsens4 (confirmed 8/25)
[Nurse] Meds managed with a pill calendar; no missed doses3
[Note to family] Please write down what time she wakes at night and what she does3
- Start
- 18:20
- End
- 18:34
- In scope
- 12
- Sent via care platform
- 9
- Sent by fax
- 2
- Not sent
- 1 (no recipient on file)
- Errors
- 0
Counts are examples. The staff member in charge checks the one that wasn’t sent and follows up.
Home care guidance report (sample)Physician reviews and finalizes
To: Kamome Care Management Office (fictional)
The passages marked in step 2 now appear in the fields with the same numbers.
Fields the robot fills in
Client: Patient A (fictional) / Visit date: Tue, Sep 8, 2026 / Medical institution: Sakura Home Care Clinic (fictional) / Diagnoses: Alzheimer’s dementia, hypertension, chronic low back pain after lumbar compression fracture
(Before entry)
1Client’s condition and clinical course
Dementia is progressing slowly, and 2–3 times a week she gets up at night and searches for things. Blood pressure is stable, and low back pain has not worsened. If the nighttime behavior continues, we will consider reviewing her oral medications.Physician added the plan
(Before drafting)
2Notes on using care services, care methods, etc.
She walks indoors by holding onto furniture, so please watch for falls. We would like to discuss increasing adult day care from twice to three times a week.Please consider whether adult day care can go up to three times a week, while keeping an eye on how tired she gets.Reworded
(Before drafting)
3Notes on daily living, social issues, etc.
She manages her medications with a pill calendar and has not missed any doses. We have asked her family to write down what time she wakes at night and what she does.
(Before drafting)
4Information on medical and long-term care at the end of life, etc.
The patient wishes to stay at home as long as possible, and her daughter agrees. If her condition suddenly worsens, she wishes to be taken to the hospital by ambulance (confirmed on August 25).
(Before drafting)
Fields 3 and 4, and the fields the robot filled in, were left unchanged.
PDFGuidance report (Patient A, Sep 8).pdf
Care platformSent to Kamome Care Management Office (fictional)
FAXSent to recipients not on the care platform
Action neededStaff follow up on the 1 unsent document
Home-visit nursing report (nurse → attending physician) – Deployed
- Patient B (fictional)To createCreated
- Patient E (fictional)To create
- Patient F (fictional)To create
- Patient G (fictional)New: nurse writes
Tue, Sep 1BP 128/74, SpO2 96%, weight 48.2 kg. Lower-leg edema (±). No shortness of breath. No missed doses
Tue, Sep 8BP 138/80, SpO2 95%, weight 49.0 kg (+0.8 kg from last week)Change. Lower-leg edema (+). Short of breath on stairs. Daughter says “he’s been eating more ready-made meals”Lifestyle. Reported to attending physician by phone; temporary diuretic increase orderedAction
Tue, Sep 15BP 126/72, SpO2 96%, weight 48.4 kg. Lower-leg edema (±). Discussed salt restriction with patient and daughter
Tue, Sep 29BP 124/70, SpO2 97%, weight 48.1 kg. No edema. No shortness of breathProgress. Walks steadily indoors using handrails. No missed doses
Home-visit nursing report (sample)Nurse reviews and edits
To: Attending physician, Sakura Home Care Clinic (fictional)
Fields the robot fills inCopied and updated
Client: Patient B (fictional), age 81 / Report date: Sep 30, 2026 / Visit dates: Sep 1, 8, 15, and 29
(Before entry)
Clinical course
On September 8, weight was up 0.8 kg from the previous week, with lower-leg edema and shortness of breath on stairs. This was reported to the attending physician, who ordered a temporary increase in the diuretic. Because the patient had been eating more ready-made meals, salt restriction was discussed with the patient and family. Weight has since stayed between 48.1 and 48.4 kg, and lower-leg edema and shortness of breath have resolvedno lower-leg edema or shortness of breath was observed at the September 29 visitStates when observed. Blood pressure has been 124–138/70–80 mmHg and SpO2 95–97%. No doses have been missed. We will continue monitoring next month, focusing on changes in weight and lower-leg edema.Next month’s plan
(Before drafting)
Nursing care details and other fields
Copied from the previous report and checked by the nurse.
The fields the robot filled in and the nursing care details were left unchanged.
FilingSeptember report filed in the EHR
PDFReport (Patient B, Sep).pdf
Patient listPatient B: Created
SendingAfter the nurse reviews and edits
Home-visit nursing order (attending physician → home-visit nursing station) – In progress
A month of records (excerpt)Last month’s orderPatient list
1. Items for the physician to check (5)
- Level of independence in daily living is estimated from the records.
- Pressure ulcer depth is left blank because it will be assessed at the next visit.
- Facility name, physician name, issue date, and recipient are left blank for the physician to fill in and seal.
- IV infusion orders are left blank because the medical record doesn’t mention any.
- Names of other home-visit nursing stations are unknown because the medical record doesn’t state them.
2. Medical record entries the draft is based on
#1 Type 2 diabetes #2 Chronic heart failure (NYHA class III, EF 38%) #3 Prior cerebral infarction (right hemiplegia, dysphagia)
S) Wife: “The swelling gets worse in the evening. At night he seems to struggle to breathe.”
O) Weight 52.1 kg (+1.3 kg since last visit). Pitting edema in both lower legs (++)1. Redness and peeling skin over the sacrum; depth to be assessed next visitNo basis: leave blank. Tube feeding via gastrostomy. Home oxygen 1 L/min on exertion only3. Needs full assistance with transfers; can sit in a wheelchair during the day. Follows simple instructions but is disoriented to date and time2
A) Heart failure trending worse1. Suspected pressure ulcer forming on the sacrum
P) Furosemide increased from 20 mg to 40 mg (once daily, morning). Home-visit nursing asked to check weight and edema daily. Contact our clinic if weight rises 1 kg or more in 2 days. Protect the sacrum, observe at least twice a week, and teach repositioning. Continue wheelchair transfer training4. Home-visit nursing continues 3 times a week
IV orders: not recorded
Physician’s edits (sample): Dementia level IIIa → IIb; wording of “Condition and treatment status”
Home-visit nursing order (sample, part of the form)Physician reviews and signs
This is an AI-generated draft, not an official home-visit nursing order. It can be issued only after the attending physician has reviewed and edited every item and signed (sealed) it.
AI draft / Not reviewed
The passages marked in step 2 now appear in the fields with the same numbers.
Fields the robot updates
Order period: Sep 1–30, 2026 / Date written: Aug 25, 2026 / Patient: Patient C (fictional), age 87 / Current medications: furosemide 40 mg tablet, once daily in the morning; enalapril maleate 2.5 mg tablet, once daily in the morning; linagliptin 5 mg tablet, once daily in the morning
(Before update)
1Condition and treatment status
Chronic heart failure is trending worse, with weight gain and edema in both lower legs. Receiving tube feeding via gastrostomy. Suspected pressure ulcer forming on the sacrum.Redness and peeling skin over the sacrum (depth to be assessed next visit).Reworded to match findings Needs full assistance with transfers; able to sit in a wheelchair.
(Before drafting)
2Level of independence in daily livingNeeds review
Bedridden level: J1 J2 A1 A2 B1 B2 C1 C2
Dementia level: I IIa IIbEstimate corrected IIIa IIIb IV M
Unchanged from last month (filled in by the physician on the first order)
Primary diagnoses: type 2 diabetes, chronic heart failure, prior cerebral infarction / Emergency contact: same as last month’s order
3Medical devices in use
Oxygen (1 L/min, on exertion) Tube feeding (gastrostomy)
Pressure ulcer depth
BlankRecord: “assess next visit”
4Notes and instructions
I. Notes on guidance for daily living: Check weight and edema daily. Contact our clinic if weight rises 1 kg or more in 2 days. / II-1. Rehabilitation: Wheelchair transfer training. / II-2. Pressure ulcer care: Protect the sacrum and observe at least twice a week. Teach repositioning.
(Before drafting)
IV infusion orders
BlankNot in the record
PDFHome-visit nursing order (AI draft).pdf
Save and printSaved to the designated folder; 1 copy printed
Review sheetItems to check: 5
IssuingAfter the physician signs (seals) it
RPAFields and actions the robot handlesAIRecord excerpts and draftsPersonReview and editDashed fields stay blank because there is no basis for them in the record
Home care guidance report, step 1: The robot opens the EHR and pulls the notes from that day’s home visits. Along with the physician’s entries, it reads the nurse’s entries and the notes to the family.
Home care guidance report, step 2: AI reads the notes and picks out what belongs in each of the guidance report’s four fields. At this stage it also spots typos like “pian” and abbreviations like “day svc” and “CM.”
Home care guidance report, step 3: AI drafts each field, fixing typos and spelling out abbreviations in full. The robot fills in fields such as the client’s name, visit date, and diagnoses.
Home care guidance report, step 4: The physician checks the draft, adjusts the wording, adds any plans that weren’t in the notes, and finalizes the report.
Home care guidance report, step 5: The robot turns the finalized guidance report into a PDF and sends it through the regional care-coordination platform. Recipients not on the care platform get a fax, and a staff member follows up on anything that couldn’t be sent.
Home-visit nursing report, step 1: The robot reads the list of patients due for reports and pulls all of this month’s nursing records for each one. Nurses write the reports for new clients themselves.
Home-visit nursing report, step 2: AI reads the records from all four visits and picks out the key points of the month, such as changes in weight and edema, and what was reported to the attending physician and ordered in response.
Home-visit nursing report, step 3: The robot copies last month’s report and updates the report date and visit dates. In the “Clinical course” field, AI enters a draft written in the station’s own style.
Home-visit nursing report, step 4: The robot files the draft report in the EHR, exports a PDF, and marks the patient “Created” on the patient list.
Home-visit nursing report, step 5: The nurse checks the draft, notes when findings were observed, adds the plan for next month, and then sends the report.
Home-visit nursing order, step 1: Overnight, the robot reads the list of eligible patients and pulls last month’s order and a month of medical records for each.
Home-visit nursing order, step 2: AI reads the records and picks out what covers the patient’s condition and treatment, along with anything that amounts to instructions for home-visit nursing. It also identifies items the record gives no basis for, such as pressure ulcer depth.
Home-visit nursing order, step 3: AI drafts “Condition and treatment status” and other fields, leaving any field without a basis blank. The robot updates the order period, age, and current medications.
Home-visit nursing order, step 4: The robot exports the PDF, then saves and prints it. The document carries a note that it is an AI-generated draft, along with a review sheet listing the items for the physician to check.
Home-visit nursing order, step 5: In the morning, the physician works through the review sheet item by item, corrects estimated items and wording, then signs (seals) and issues the order.
How this diagram works
Each time you choose a document or a step, the same mock-up screen switches what it shows. All of the text is in the page itself. The animation plays only when the mock-up comes into view, and it stops at the last step.
State: radio + :has() / Previous and Next: the label’s for attribute / Typing effect: clip-path and steps() / Autoplay: a JavaScript timer (only while on screen, via IntersectionObserver) / Switching documents: View Transitions
On devices set to reduce motion, or where scripts don’t run, clicking a step shows the same content.
*This is a sample for illustration. All medical institutions, care agencies, patients, records, and document text are fictional. Times, forms, who handles which fields, and when reviews happen are all decided together with you, after we learn how your facility works.
Home care guidance reports and home-visit nursing reports are already in operation at client sites. Support for home-visit nursing orders is currently in progress.
Talk to us about preparing documents
CASES
How the steps and work time changed after adoption
For three tasks at client sites, we show the steps before and after adoption side by side, along with the change in work time. Flip the switch to see the work people used to do move over to the robot and AI.
Steps done by people
Before652After221
Home-care clinic | Admin staff | Every day
Time spent drafting and sending (per month)
*Estimate based on interviews at one client facility. Results vary with the number of visits and how work is organized.
Staff member at a client site (paraphrased)Reports that used to go out late are now sent the same day.
Home-visit nursing station | Nursing team | Month-end
Time spent writing reports (month-end)
*Results at a client site. Actual results vary by facility, task, and workflow.
The time still needed goes mainly to reviewing and touching up drafts.
Staff member at a client site (paraphrased)There’s less variation in writing time and quality from one nurse to the next.
Clinics and hospitals | Admin staff | Every month
Time spent tallying and preparing submission data (per month)
*Results at a client site. Actual results vary by facility, task, and workflow.
Staff member at a client site (paraphrased)With less tallying by hand, it’s easier to make time to check the data before we submit it.
Home care guidance report: before
Home care guidance report: after
Home-visit nursing report: before
Home-visit nursing report: after
Extracting and compiling outpatient data submission add-on items: before
Extracting and compiling outpatient data submission add-on items: after
Home care guidance report, before: Medical administrative staff did all six steps by hand, from transferring the notes and exporting the PDF to sending it via the care platform and by fax.
Home care guidance report, after: The robot handles data transfer, PDF export, and sending, and AI writes the draft. People now do two things: the physician reviews and edits the draft, and staff follow up on documents that couldn’t be sent.
Home-visit nursing report, before: Nurses reread a month of records and wrote up the report themselves, then handled data entry, PDF export, and updating the patient list.
Home-visit nursing report, after: The robot retrieves the records, creates the reports, and updates the patient list, and AI summarizes the clinical course. Nurses review, edit, and send the drafts, and write the reports for new clients themselves.
Extracting and compiling outpatient data submission add-on items, before: Medical administrative staff entered the eligible items by hand every day, then checked the eligible patients at month-end and prepared the data for submission.
Extracting and compiling outpatient data submission add-on items, after: The robot retrieves the records and builds the submission list, and AI extracts the eligible items. The staff member in charge reviews the list and submits it.
How this diagram works
Cards for the same task share the same name, so when you switch views, the browser automatically animates each card from its old position to its new one.
View Transitions (view-transition-name) / radio and checkbox + :has() / CSS Grid
In browsers that don’t support this, the switch is instant, but the cards end up in the same places. In older browsers, the before and after views of all three cases are stacked vertically.
*This diagram is based on client cases. Facility names are withheld. Steps, and how they are counted, vary by facility.
Automation with MedSync is not meant to cut staff. It frees the time spent on documents and tallying for reviewing the work and caring for patients.
See more case studiesTalk to us about this task
DEMO
Watch the robot and AI at work on real screens
This short video (about 30 seconds) shows
Video chapters
- 0:02RPA① The robot (AUTORO) retrieves the month’s medical records
- 0:05RPA④ The robot sends the records to
M-Automate - 0:08AI⑤
M-Automate drafts the document to match the facility’s form - 0:11RPA⑥ The robot exports the draft as a PDF
- 0:14RPARun log listing every step from retrieval to saving (4× speed)
- 0:17AIThe exported home-visit nursing order (AI draft)
- 0:22PersonAI draft review sheet: lists the items the physician needs to check
- 0:25PersonAI produces only a draft. Healthcare professionals review and edit it and make the final decision
The numbers in the video are the robot’s step numbers. Steps that only prepare the connection are not shown.
The physician reviews and edits the content using the “AI draft review sheet” shown at the end. The screens are from the actual product (Japanese interface), and the case is fictional. Some areas, such as the sample medication fields, are blurred. The video has no sound.
AI TUNING
Tune the AI to your facility’s writing style
Draft quality depends heavily on the prompt (written instructions to the AI) and on the chart entries it works from. Even from the same records, the rules written into the prompt change the tone and format of the draft.
- Tone
- NeutralPolite (addressed to the attending physician)
- Format
- One paragraph, about 100 wordsItem by item, in short sentences
- Common rules
- Write only what is in the record
Give measurements as ranges for the month
State the date each finding was observed
Home-visit nursing reportSample
- To
- Attending physician, Sakura Home Care Clinic (fictional)
- Client
- Patient B (fictional), September 2026
Clinical courseAI draft
Neutral, paragraph
On September 8, weight was up 0.8 kg from the previous week, and lower-leg edema and shortness of breath on stairs were noted. The attending physician was notified and ordered a temporary increase in the diuretic. Because the patient had been eating more ready-made meals, salt restriction was discussed with the patient and family. Weight then stayed at 48.1–48.4 kg, and at the September 29 visit, lower-leg edema and shortness of breath were absent. Blood pressure remained at 124–138/70–80 mmHg and SpO2 at 95–97%. There were no missed doses.
Neutral, bullet points
- Weight: Up 0.8 kg from the previous week on September 8. Stable at 48.1–48.4 kg since.
- Symptoms: Lower-leg edema and shortness of breath on stairs present on September 8. Neither noted at the September 29 visit.
- Action taken: Reported to the attending physician by phone on September 8; temporary diuretic increase ordered.
- Daily life: Eating more ready-made meals. Salt restriction discussed with patient and family.
- Vitals: Blood pressure 124–138/70–80 mmHg, SpO2 95–97% throughout.
- Medication: No missed doses.
Polite, paragraph
On September 8, his weight was up 0.8 kg from the previous week, and we observed lower-leg edema and shortness of breath on stairs. We reported this to you by phone and received your instruction to temporarily increase the diuretic. Because he had been eating more ready-made meals, we discussed salt restriction with him and his family. His weight has since stayed at 48.1–48.4 kg, and at the September 29 visit, we observed no lower-leg edema or shortness of breath. His blood pressure has remained at 124–138/70–80 mmHg, and his SpO2 at 95–97%. He has not missed any doses.
Polite, bullet points
- Weight: On September 8, his weight increased by 0.8 kg from the previous week. It has since stayed at 48.1–48.4 kg.
- Symptoms: On September 8, we observed lower-leg edema and shortness of breath on stairs. At the September 29 visit, we observed neither.
- Action taken: On September 8, we reported this to you by phone and received your instruction to temporarily increase the diuretic.
- Daily life: Because he had been eating more ready-made meals, we discussed salt restriction with him and his family.
- Vitals: His blood pressure has remained at 124–138/70–80 mmHg, and his SpO2 at 95–97%.
- Medication: He has not missed any doses.
NurseReviews and edits the draft, then sends it to the attending physician.
Underlined parts change depending on the writing rules.
Original nursing recordsSame for every style
Tue, Sep 1BP 128/74, P 76, SpO2 96%, weight 48.2 kg. Lower-leg edema (±). No shortness of breath. Medications checked, no missed doses
Tue, Sep 8BP 138/80, P 84, SpO2 95%, weight 49.0 kg (+0.8 kg from last week). Lower-leg edema (+). Short of breath on stairs. Daughter: “He’s been eating more ready-made meals.” Reported to attending physician by phone; temporary diuretic increase ordered
Tue, Sep 15BP 126/72, P 78, SpO2 96%, weight 48.4 kg. Lower-leg edema (±). Discussed salt restriction with patient and daughter
Tue, Sep 29BP 124/70, P 74, SpO2 97%, weight 48.1 kg. No edema. No shortness of breath. Walks steadily indoors using handrails. Assisted with bathing; no skin problems. Medication checked; no missed doses
Neutral, paragraph: Sums up the month’s course in a paragraph of about 100 words, using only facts from the record.
Neutral, bullet points: The same content, split into items such as weight, symptoms, and action taken, and written in short sentences.
Polite, paragraph: Written as a report to the attending physician, in a polite, respectful tone.
Polite, bullet points: Split into items, then written in polite sentences addressed to the attending physician.
How this diagram works
All four drafts are written into the page, and only the one that matches your choices is shown. The browser automatically animates the change from one draft to the next.
radio + :has() / View Transitions (view-transition-name) / CSS Grid
In browsers that don’t support this, the switch is instant. In older browsers, the four drafts are listed one after another.
*Sample for illustration. The records and drafts are fictional. We ask each facility what rules it wants, agree on them together, and build them into the AI prompt.
Drafts are more consistent when chart entries follow a fairly uniform style.
HOW IT WORKS
The robot works the screens, and AI handles the writing
RPA (robotic process automation) uses software robots to carry out the routine operations people would otherwise perform on a computer.
Diagram of the path records take. The robot pulls records from the EHR and passes them to
- 1RPAThe robot retrieves the items and records it needs from the EHR and business systems.
- 2AI
M-Automate reads the records according to your facility’s rules, extracts and summarizes the information needed, and creates draft documents and lists. - 3PersonThe physician, nurse, or other staff member in charge checks the draft and any points that need individual attention, edits as needed, and finalizes it.
- 4RPAThe robot transfers and files the finalized text, then handles PDF export, saving, sending, and printing.
The robot retrieves the records and passes them to
This is the basic flow. For some tasks, the draft may be filed or exported to PDF before a person reviews it. Before anything is sent outside your facility, a person reviews and edits the content. *Configuration depends on your environment.
-
Facilities already using RPA
You can add document drafting and summarization to robots already in operation.
-
Facilities not yet using RPA
We help you design and develop robots built to work with AI from the start. MedSync is an authorized reseller of BizRobo!, an RPA tool.
-
Supported RPA tools
M-Automate can be built into BizRobo! and AUTORO robots, with no limit on the number of robots. -
With your current EHR
RPA operates the screen just as a person does. That means even if your EHR has no API (a way for systems to exchange data), RPA may still be able to retrieve records and transfer data. We’ll check your environment and let you know whether it can be supported.
When the robot and AI runSample
- After the day’s visitsThe day’s chart entries are all in
- OvernightThe robot collects the records, and AI writes the drafts
- Next morningPhysicians and nurses review and edit
- After reviewThe robot exports PDFs and sends them
Run times and review timing are set to fit each task and your facility’s workflow.
RPA and AI processing can run in batches at times you set, including overnight.
SAFETY
Designed around human review and proper handling of medical information
Text written by AI may contain errors or content that is not in the medical record. The physician, nurse, or other staff member in charge reviews and edits it before it is finalized.
1Define the AI’s scope before it goes live
Left to AI
- Fix obvious typos and missing words
- Expand abbreviations into full terms
- Turn conversational phrasing into written style
Decided by your facility
- How much to polish the wording
- Whether the physician’s wording may be changed
- Who does the review
Not allowed
- Writing anything that is not in the record
- Skipping human review
The agreed scope is built into the AI prompt.
2How AI and people share the work
- We instruct the AI to write only from information in the medical record.
- Physicians, nurses, or staff in charge review and finalize the content.
- Documents that are hard to standardize, such as those for new patients, are written by people.
3Handling medical information
- Processing takes place in your facility’s dedicated environment on servers in Japan. It can also be deployed on a closed network that is not connected to the internet.
- Client certificates and IP restrictions limit where connections can come from, and we log who did what, and when.
- Neither your inputs nor the generated output is used to train AI models.
- Designed in line with government guidance on medical information security, including the Guidelines for the Security Management of Medical Information Systems from Japan’s Ministry of Health, Labour and Welfare (MHLW).
*Configuration depends on your environment.
RPA
Tasks that need only screen operations can be automated with RPA
Rule-based tasks such as data entry, data transfer, tallying, and sending can sometimes be automated with RPA robots alone. When a task also involves reading records and turning them into text, we add
RPARPA implementation and support
Rule-based screen operations
- Open the EHR
- Search for the patient
- Retrieve the records needed
- Enter into Excel
- Save PDFs
- Enter data into another system
- Log the results
AI
Reading records and turning them into text
- Read records
- Find key points
- Summarize
- Write drafts
This is the product featured on this page.
Combined, they run the whole process as one flow: retrieving records, drafting, and sending after human review.
- RPARetrieve records
- AIDraft
- PersonReview and edit
- RPASend
FAQ
Frequently asked questions
Could the AI write something incorrect?
Text written by AI may contain errors or content that is not in the medical record. That’s why AI output is treated as a draft, and the physician, nurse, or other staff member in charge reviews and edits it before it is finalized. We instruct the AI to write only from information in the medical record.
Where is patient information processed?
Records are processed in your facility’s dedicated environment on servers in Japan. This can also be deployed on a closed network that is not connected to the internet. Client certificates and IP restrictions limit where connections can come from, and we log who did what, and when. Configuration depends on your environment. Neither your inputs nor the generated output is used to train AI models.
Won’t this actually add to physicians’ review work?
We keep the physician review step in place. Review does take time, but to keep corrections to a minimum, we tailor the prompts to your facility’s writing style, and we ask which parts get corrected most so we can refine them. Review timing is also set to fit your facility’s workflow. For example, drafts can be prepared overnight for review the next morning.
Clients using
Can we use it with our current EHR?
Replacing your EHR is not a prerequisite. RPA operates the screen just as a person does. That means even if your EHR has no API (a way for systems to exchange data), RPA may still be able to retrieve records and transfer data. We’ll check your systems and network environment and let you know whether it can be supported.
Can we use it if we don’t have RPA?
Can hospitals use it too?
What happens if a robot stops?
A robot can stop when, for example, an EHR password expires or a screen changes. Run results are logged, so you can see which documents weren’t processed and which recipients weren’t reached. Modifications needed after changes to screens or in-house procedures are covered by a maintenance and improvement contract. We ask your staff member in charge to handle the daily runs and check the results.
When will we know the expected ROI?
We’ll show you before development begins. We ask which tasks you want to automate and how long they take today, then walk you through the expected time savings together with the costs.
CONTACT
Tell us which documents and tasks are taking up your time
Whether it’s preparing documents or monthly tallying, we’ll go through your current steps with you and sort out what
Related servicesRPA implementation and support
Information as of September 2026.
