Superintelligence Prep for Urology Offices: Referrals and Procedure Prep
How a urology office can use an AI agent to track referral intake, send approved prep instructions and follow-up, keeping clinical questions with staff.

An AI agent for a urology office can track each referral from arrival to first visit, record which approved prep instructions were sent and acknowledged, and prompt staff when a follow-up is overdue. It does not interpret symptoms, change medication instructions or decide whether a procedure is appropriate, because those decisions stay with clinicians and the staff they authorize. The useful job is administrative memory and reminders, bounded by written approvals.
Key takeaways
- Referral intake is a tracking problem before it is a clinical one: the agent can record what arrived, what is missing and who owns the next step.
- Prep instructions should be staff-approved text that the agent sends unchanged, never generated answers to patient questions.
- Any question about symptoms, medications or whether to proceed routes to a named staff member with a clear handoff.
- Follow-up tasks need an owner, a due point and a record of completion, reviewed by the office on a set schedule.
- Start with one referral stream and one success criterion, then widen permissions only after the office reviews the results.
Where referral intake loses time
Referrals reach a urology office by fax, portal, phone, email and sometimes through the patient, and each channel carries different pieces of information. The clinical question is rarely the hardest part. The slow part is memory: which referral is missing a form, which attachment has not arrived, and who was supposed to call the referring office back. An agent that keeps one intake record per referral gives the front desk and clinical staff a single place to see those open items.
A usable intake record usually holds these fields:
- The referring clinician and a return contact for letters and results
- The reason for referral, quoted from the referring clinician's note
- Attached records and imaging, with the date each one arrived
- Coverage details and whether an authorization request is needed
- The patient's preferred contact method and language
- The named owner of each open item and when it was last touched
The agent reads incoming documents, fills the record where the source text is clear, marks what is missing, and drafts the request to the referring office for a staff member to approve. It does not judge urgency or decide whether a patient should be seen sooner. That judgment belongs to clinicians, and the record should show when it was made and by whom. Offices that run intake for other clinical services face the same limit, which is why the intake and scheduling limits in behavioral health clinic intake make a useful comparison.
Illustrative scenario: the referral that stalls
Illustrative scenario: A primary care office faxes a referral late on a Friday. The form is complete, but the recent laboratory results it mentions are not attached. The front desk assumes they were sent with the fax and books a consultation for the following week. Nobody notices until the patient arrives and the chart has no results to review, so the visit is partly wasted and the patient must return.
With an agent-supported intake record, the missing laboratory item appears as an open item as soon as the fax is processed. A staff member receives a drafted request to the referring office, the consultation shows as pending records, and the scheduling step cannot be marked complete until someone closes the item or the clinician accepts the visit without it. The agent made no clinical decision. It made the gap visible and kept it visible until a person resolved it.
Writing prep instructions without giving clinical advice
Procedure preparation is where the boundary matters most. Instructions about fasting, medication timing, hydration, bringing a companion or arriving with specific preparations are clinical content, even when they read like logistics. An agent should never write its own version of them, paraphrase them to fit a patient's question, or infer what applies to a particular medication. The safer design is a library of instruction sheets that the office's clinicians approve, each tied to one procedure type, versioned, and given a named owner and review date.
Once that library exists, the agent's job is narrow: send the approved sheet unchanged, log that it was sent and when, and ask the patient to confirm receipt. Patient messages are untrusted input, and they can contain text that looks like instructions to the agent, so the agent follows the office's rules rather than whatever a message asks for. The OWASP prompt injection guidance explains why this separation between instructions and user content matters.
When a patient writes in, a simple decision guide keeps the agent inside its limits:
- The message asks for the approved prep sheet for a booked procedure: send that sheet, log it, and request confirmation.
- The message asks about a specific medication, a symptom, or whether to go ahead: do not answer. Create a staff task with the patient's words attached and tell the patient a staff member will reply.
- The message asks about logistics such as parking, arrival time or what to bring: answer only from the approved logistics sheet.
- The message describes a warning sign or an emergency: send the office's approved emergency wording and alert the on-call staff member at once.
- The message fits none of the above: hold the reply, notify staff, and do not guess.
A step-by-step plan for the first agent role
Start with one role, not the whole front office. A workable first engagement follows this order:
- Pick one referral stream, such as referrals from a single primary care group, so the intake rules can be tested against familiar paperwork.
- Write the intake checklist with the front desk lead and a clinician, and agree on which fields must be complete before a consultation is booked.
- Build the prep sheet library for the procedures the office performs most often, with an owner, version and review date on each sheet.
- Grant read access to the referral inbox and write access only to the intake record and staff tasks. The agent should not edit clinical fields or send free-form messages.
- Define the success criterion in plain terms, for example: every referral in the stream has either a complete intake record or an open item with a named owner, checked each morning.
- Run the agent in shadow mode, where it flags items but staff act as they did before, and compare its flags with what staff found.
- Review every miss with the office, adjust the rules, and widen permissions one step at a time after each review is signed off.
Follow-up, records and access after the visit
Follow-up is where many offices lose the most ground, because the work is spread across clinicians, schedulers, billing and the referring office. The agent can track what is due and who owns it: a return visit that needs booking, a pending result, a letter back to the referring clinician, or a post-procedure check-in that a nurse has scheduled. It can send reminders using approved wording and record each completed step. The recall logic used in other clinical intake work, such as a chiropractic intake and recall workflow, applies here too: the agent keeps the list, and people close it.
It should not decide when a result is ready to share or what it means. Result communication stays with the clinician who ordered the test, and the referring letter goes out only from a staff-approved template after the clinician signs off.
Access should follow least privilege. Every action the agent takes should map to a defined role, be logged, and be reviewed on a schedule, especially when staff change roles. The OWASP authorization guidance is a practical checklist for this work. Patient data handling must follow the privacy rules that apply to the office and the places where its patients live, so involve the office's compliance advisor before connecting any system that holds patient health information. The NIST AI Risk Management Framework gives a structure for recording risks and checking them over time.
How Be Human's managed approach fits a urology office
Be Human's first engagement defines one useful role, the memory and tools it needs, the approval boundaries, and an observable success criterion, which for a referral agent is usually the state of the intake record and the follow-up list. Managed operations then keep the agent working as the office changes: monitoring the queues, running recurring evaluations against the prep library and escalation rules, maintaining connections to the referral inbox and scheduling system, reviewing permissions when staff roles change, and handling incidents when something goes wrong. Service is remote and available in English and Portuguese. The office keeps the clinical decisions, and Be Human keeps the agent inside the boundaries those decisions set. If you want to find the first task in your office worth handing to an agent, take the free Be Human AI diagnostic.
Frequently asked questions
Can an AI agent answer patient questions about procedure prep? It can send the prep sheets your clinicians have approved, unchanged, and log that they were sent. It should not write its own explanations of fasting, medications or what to do on the day. Questions outside the approved material go to a staff member with the patient's message attached.
Does an agent decide which referrals are urgent? No. Urgency triage belongs to clinicians, and the agent's part is to make sure each referral arrives complete and visible. It can flag a missing item or a time-sensitive note in the referral text so a person sees it quickly, but the flag does not set priority.
Where should a urology office start? Start with one referral stream and one success criterion, such as every referral having either a complete record or a named open item. Build the prep library and the handoff rules before connecting systems, and run the agent in shadow mode before it takes any action on its own.
Does this replace the front desk? No. The aim is that staff spend less time searching for missing documents and chasing open items, and more time on the calls and decisions that need a person. Staff still approve outgoing messages, own the open items and review the agent's work on a regular schedule.
Your next step
See where an AI agent fits in your company first.
Take the free Be Human AI diagnostic. A few questions about your work show which task to hand to an agent first, what it needs to know and where a person approves.
Be Human editorial guidance. Examples are illustrative unless explicitly attributed. Sources provide background; implementation choices depend on your workflow.


