AI for Podiatry Practices: Referral Tracking and Recall Reminders

An AI agent for a podiatry practice can track referrals from arrival to a returned consultation report and send recall reminders for patients due for follow-up, all within rules the podiatrist writes. It handles the administrative layer: missing paperwork, routing, scheduling requests, overdue lists and reminder messages. The podiatrist keeps every diagnosis, triage decision and treatment plan, and the agent is built so it cannot make those decisions.

Key takeaways

• Referral tracking is mainly a status problem: every referral needs a named owner, a current state and a due date before any software is involved.
• Recall intervals should come from the podiatrist's written protocol, not from a default the software chose.
• The agent may file, route, remind, flag and draft. It should not assess urgency, interpret symptoms, recommend treatment or change a plan.
• Start with one referral source and one recall type, then widen only after the first one runs without unresolved exceptions.
• Every agent action should leave a record a clinician can review, and its access should be limited to what its role needs.

Where the gaps usually sit in a podiatry office

Referrals arrive from family physicians, diabetes clinics, orthopedic teams and sports medicine colleagues, often by fax, portal message, phone call or letter. Each channel has its own format, and the information needed to act on a referral is frequently split between them. The consultation report that should go back to the referrer may wait in a queue until someone has time, and the referrer may not learn the outcome for weeks. On the patient side, follow-up depends on remembering who is due for a check, a footwear reassessment or an orthotic review, and that memory often lives in a spreadsheet, a calendar color or one staff member's head.

These gaps are administrative. They are also where patients slip out of care, which is why they matter clinically even though no one makes a clinical decision when a letter sits unsent.

Illustrative scenario: a family physician sends a referral for a patient with a nail problem and reduced sensation in both feet. The agent logs the referral, checks that the referring note and the medication list are attached, asks the front desk to request the missing laboratory report, and places the file in the podiatrist's review queue. It does not rank how urgent the case is. The podiatrist reads the note, chooses the appointment type and decides the plan.

How to track a referral from arrival to a closed loop

A referral tracker works best when every referral moves through a short, fixed set of states. The list below is a common starting point. Adjust the names to match your practice, but keep ownership clear.

• Received: source, date and referring clinician recorded
• Complete: referral note and required documents present, or missing items requested
• Clinician review: waiting for the podiatrist to accept, decline or ask for more information
• Scheduled: appointment booked with the type the podiatrist selected
• Seen: visit completed and clinical note filed
• Report returned: consultation letter sent to the referrer
• Closed: report confirmed as delivered, or referral declined with a recorded reason

The closed loop is the part most often missing. A referral that reaches seen but never reaches report returned leaves the referrer without the information they need for their own patient. The agent's job here is to hold each file between states and surface any referral that has stayed in one state longer than the threshold the practice sets. That threshold is a management decision, and the owner of each state should be named on paper before the agent starts.

The same logic applies to deadline-driven work outside healthcare. The intake and deadline setup described in AI Agents for Law Firms: Setup Guide for Intake and Deadlines follows the same principle: the software watches the clock, and a named person owns the decision.

Recall reminders: who sets the interval and what gets sent

Recall is where many practices lose the thread. Patients who are at risk for foot complications, or who are on an active orthotic or footwear plan, need periodic review, and the interval belongs to the podiatrist and the protocol the clinic follows. The agent does not choose that interval. It calculates due dates from the interval stored on the patient record, sends reminders through approved channels and produces a daily list of overdue patients for staff to call.

Recall logic in other clinical scheduling follows the same pattern. AI Agents for Orthodontists: Aligner Tracking and Recalls shows how recall tracking and patient messaging can stay separate from the treatment decisions that create the recall in the first place.

A short checklist for recall setup:

• Each recall type has an interval written by a podiatrist and signed off by the clinic owner
• Reminder wording is approved once by the clinical lead and reused without improvisation
• Patients who ask to reschedule or cancel are routed to staff, not answered with advice
• Any reply that mentions pain, a wound, a color change or a new symptom goes to a human at once, with no automated answer
• Overdue recalls appear on a staff list every working day, so nothing is dropped silently
• Opt-outs and language preferences are honored and recorded

Reminders should be written in the patient's preferred language. Many practices serve patients in more than one language, and a reminder that is accurate but unreadable does not do its job.

Where the podiatrist keeps every decision

The boundary is the most important design choice in this work. Clinical judgment is not a task the agent performs with slightly lower accuracy than a human. It is a task that stays with the licensed clinician by design, along with responsibility for it. The agent works around that judgment and never inside it.

A decision guide for the first version:

• The agent may file, route, remind, flag missing paperwork, draft an acknowledgment that contains no clinical content and list overdue items
• The agent may draft for review: referral reply letters, recall letter text and scheduling requests that a person will approve
• The agent must not assess urgency, interpret symptoms, recommend treatment, accept or decline a referral, or change a recall interval
• Staff must handle any patient message with clinical content, and any exception to a written rule

Referral documents come from outside the practice and may contain text that looks like instructions to a software system. A faxed or emailed note should be treated as data to be read, never as a command to be followed. The OWASP guidance on prompt injection prevention covers this class of risk. The agent's permissions should also be narrow enough that even a successful manipulation cannot send messages or change records beyond its role.

A first-month plan your team can run

The first deployment should be small enough that a mistake is visible and cheap to correct. The sequence below keeps the podiatrist in control at each step.

• Choose one referral source and one recall type that already produce visible gaps
• Write the referral states on one page and name the owner of each state
• Have the podiatrist write the recall intervals and the boundary rules, also on one page
• Grant the agent read access to the referral inbox and the calendar, and write access limited to drafts and status fields
• Run the agent alongside staff for a period, comparing its queue with what staff actually did by hand
• Review every exception with the clinical lead and update the rules wherever the agent and the staff disagreed
• Add a second source or recall type only after the first runs without unresolved exceptions

Be Human's managed approach follows this sequence. An engagement defines one useful role, the memory and tools it needs, the approval boundaries and a success criterion the practice can observe, such as every referral reaching a closed state within the window the clinic has chosen. After go-live, monitoring, recurring evaluations, permission reviews and incident response keep the agent inside those boundaries as the practice changes. Be Human designs and operates this infrastructure remotely, and it does not replace the clinical judgment described above.

Access, privacy and records you can defend

Patient information in a podiatry practice is sensitive, and the rules that apply depend on where the practice operates. In Brazil, the Lei Geral de Proteção de Dados governs personal data, and the Autoridade Nacional de Proteção de Dados is the national authority to consult. Practices should confirm their own obligations with their regulator and legal adviser before connecting any system to patient records.

Access should follow the principle of least privilege. The agent receives only the fields its role needs, and each permission is reviewed on a schedule rather than only when someone remembers. The OWASP authorization guidance is a useful reference for checking that each action is allowed for the role requesting it, instead of relying on the interface to hide options.

Every agent action should leave a record: what was read, what was drafted, what was sent, when, and which staff member or rule approved it. Those records let the podiatrist answer a question about a referral months later without reconstructing it from memory. A practice that cannot see what its tools did cannot judge whether they are working. For broader governance questions, the NIST AI Risk Management Framework offers a shared vocabulary for mapping, measuring and managing risk in AI systems.

For practices that want to see which referral or recall task would be the right first one to hand to an agent, the free Be Human AI diagnostic is a plain place to start.

Frequently asked questions

Can an AI agent decide which referrals are urgent? No. It can check whether a referral is complete, route it to the podiatrist's queue and flag items that have waited too long, but ranking urgency is a clinical judgment. The podiatrist or a designated clinician reads each referral and sets the next step, and the agent's rules state that explicitly.

Do recall reminders replace the clinical follow-up plan? They do not. A reminder only keeps a scheduled review from being forgotten. The content of the review, any change to the interval and any response to a patient's symptoms remain with the podiatrist and the clinical team.

How do we know the agent is doing its job? Compare its queues with a sample of what staff handled by hand, track referrals that stay in one state past the threshold you set, and review every exception with the clinical lead. Those checks show whether the boundaries are working or need adjusting.

Is this a reasonable starting point for a small practice? Often it is, because one referral source and one recall type are enough to show whether the approach fits. The scope stays small, the podiatrist writes the rules, and the practice decides later whether to widen the role.

Be Human editorial guidance. Examples are illustrative unless explicitly attributed. Sources provide background; implementation choices depend on your workflow.