Sample AI Document Report: What a Referral Packet Summary Looks Like

About this example: real reports contain patient information and are covered by HIPAA and client confidentiality, so the report below is a mock-up with a fictional patient and synthetic documents. It shows the format our PDF processing workflow is designed around; actual reports are configured with each hospital.

Illustrative example · fictional patient · synthetic documents

Document packet summary

Inbound referral packet — 38 pages, received by fax

Patient

Test Patient A (fictional)

Record ID

SYN-000000

Status

Pending human review

1. Document inventory

Pages	Document type	Notes
1–2	Fax cover sheet	Sender: referring clinic
3–6	Referral letter	Reason for referral on p. 3
7–15	Progress notes	3 visits
16–21	Lab results	Most recent panel on p. 16
22–29	Imaging report	1 study
30–34	Medication list	—
35–38	Insurance card & authorization form	Form on p. 37

2. Extracted fields

Referring provider

Dr. Example Provider p. 3

Reason for referral

Evaluation of recurring chest discomfort p. 3

Requested service

Outpatient cardiology consult p. 4

Stated urgency

Routine p. 4

Insurance plan

Example Health Plan p. 35

3. Flags for the reviewer

• Missing signature on the referral letter p. 6
• Authorization form incomplete — member ID field blank p. 37
• Lab panel older than 90 days — confirm whether new labs are needed p. 16

4. Draft summary & suggested routing

Referral for outpatient cardiology evaluation of recurring chest discomfort. Packet includes three progress notes, a lab panel, and one imaging study. Two administrative gaps (signature, authorization form) should be resolved before scheduling.

Suggested queue Cardiology intake — reviewer confirms before routing

Processed on-premiseNo external API callsAudit log ID: EX-0000Reviewer: not yet assigned

What each section is for

• Document inventory — a page-by-page map of a long packet, so staff can jump straight to the part they need.
• Extracted fields — the details that drive the next step, each with a page citation the reviewer can check.
• Flags for the reviewer — missing signatures, incomplete forms, and stale results surfaced before they cause a delay.
• Draft summary and suggested routing — a starting point, never a final decision. A person confirms where the packet goes.
• Audit footer — where it was processed, that no external API was called, and who reviewed it.

How a report like this is produced

Every GofarAI workflow tool follows the same four steps on hospital hardware: read each page, extract the relevant fields, classify documents and issues, and draft an output for a human to approve. The model runs on the on-premise GofarAI platform, and every inference is logged.

Working prototype

PDF batch processing is a working prototype available for hospital pilot evaluation. The report format is configured with each pilot hospital — fields, flags, and routing queues differ by department.

See all workflow tools · How audit logging works · On-premise vs cloud AI vendors

Frequently asked questions

Is this a real AI report from a hospital?

No. It is an illustrative mock-up built for this page with a fictional patient and synthetic documents. Real reports contain protected health information and stay inside each hospital’s environment, so we never publish them.

What should an AI document report include?

At minimum: an inventory of what is in the packet, the key fields extracted with page citations so a reviewer can verify them, flags for anything missing or inconsistent, a draft summary, and a clear record of who reviewed it and when.

Why does every extracted field cite a page number?

So the reviewer can check the source in seconds instead of rereading the packet. A field without a citation is hard to verify, and anything that cannot be verified should not be trusted.

Does the AI route documents automatically?

In this design, no. The report suggests a queue; a person confirms the routing. Every suggestion and every human decision is logged.