From Medical Voicemail to Action Plan: How I Used AI to Navigate a Prescription Denia

From Medical Voicemail to Action Plan: How I Used AI to Navigate a Prescription Denial

Bottom line: Artificial intelligence did not diagnose my glaucoma, choose my medication, or argue with my doctor. It helped me do something both simpler and immediately useful: understand a medical voicemail, organize a confusing insurance problem, and turn it into a professional message that the ophthalmology office could act on.

That may not sound like the science-fiction version of medical AI. There was no robot surgeon, digital miracle cure, or glowing hologram announcing that healthcare had been reinvented before lunch. There was a cellphone voicemail, two glaucoma medications my new insurance would not cover, a pharmacy that could not send me the denial notices, and a patient trying to figure out what to do next.

In other words, it was a normal day in American healthcare.

AI did not make the medical decision. It helped move the right information through the obstacle course.

The Podcast Asked What Medical AI Will Become

In our recent Deep Dive AI podcast about artificial intelligence and medicine, we explored the larger future: AI systems that may help clinicians interpret complex records, recognize patterns, reduce administrative work, monitor patients, and support better decisions. Those possibilities deserve serious attention, careful testing, and strong safeguards.

But there is another side of medical AI that receives less attention. It is already available to ordinary people, and it does not require a hospital to purchase a million-dollar platform. AI can act as a communication and organization assistant for the messy space between the patient, doctor’s office, pharmacy, insurer, and pharmacy benefit manager.

That space is where good care can stall—not necessarily because anyone is careless, but because every participant has a different piece of the story. The pharmacy sees a rejection. The insurance company has coverage rules. The doctor knows the treatment history. The office staff manages prior authorizations. The patient receives a voicemail and somehow becomes the temporary project manager.

Listen to the Deep Dive AI podcast while you read:

My Real-World Test: Two Denied Glaucoma Medications

I take prescription eye drops for glaucoma. After an insurance change, coverage was denied for Rhopressa and Vyzulta. I remembered previously trying medications such as Lumigan and latanoprost, but I did not want to rely on memory alone or accidentally present an incomplete treatment history as fact.

I contacted the pharmacy. The pharmacy confirmed the coverage problem, but it could not forward the denial notices directly to me. The ophthalmology office could request the necessary information. Later, I received a voicemail from the office explaining the next step and providing contact details.

That voicemail contained the bridge between “my medicine was denied” and “the office has what it needs to help.” The challenge was capturing it accurately.

This is where I used Codex through ChatGPT as a working assistant. I supplied the voicemail recording and asked the AI to help me process it. The tool transcribed the message, separated the essential facts from conversational filler, identified the staff member’s contact information, and helped draft a clear email response.

The result was not a medical recommendation. It was a better handoff.

What the AI Actually Did

The workflow was straightforward:

1. Captured the source: I used the actual voicemail rather than depending on what I thought I remembered.
2. Created a transcript: The AI converted the audio into text that I could inspect.
3. Extracted action items: It identified who contacted me, why they called, what information was requested, and how I could respond.
4. Organized the medication issue: It kept the denied drugs, prior medication trials, pharmacy limitation, and requested next steps in separate, readable sections.
5. Drafted a professional email: It helped me communicate without sounding angry, confused, or as if I were telling the doctor how to practice medicine.
6. Kept me in control: I reviewed the transcript, corrected anything uncertain, and decided what to send.

The last step is the most important. AI-generated medical communication should be treated as a draft until a person checks it against the original voicemail, prescription labels, portal messages, pharmacy information, and their own records.

The AI Was a Translator Between Systems

Healthcare communication often fails because information is fragmented. A patient may understand the problem emotionally—“I cannot get the drops that control my eye pressure”—but the office needs operational details:

• Which medications were denied?
• Which pharmacy received the rejection?
• Has the insurer specified a covered alternative?
• Which medications were previously tried?
• Were they ineffective, poorly tolerated, or changed for another reason?
• Is a prior authorization or appeal needed?
• Does the patient need temporary samples or an interim plan?
• When should eye pressure be rechecked?

An AI assistant can turn a long, anxious explanation into those categories. It can also help the patient ask questions instead of making demands. “What is the treatment plan while authorization is pending?” is more useful than “The insurance company has to approve this immediately.”

The tool becomes a kind of administrative translator. It converts the patient’s story into a format that busy staff can scan and route. That does not guarantee approval, but it reduces the chance that the next step is delayed because the request is vague or missing basic information.

How Other Patients Can Use This Method

You do not need to be a programmer. You need a source, a careful prompt, and a review process.

1. Start With the Original Information

Use the actual voicemail, letter, pharmacy message, visit summary, or insurer notice whenever possible. If you only have your memory, say so clearly. Never ask the AI to fill in a missing drug name, dosage, date, diagnosis, or test result.

2. Remove Information the AI Does Not Need

Before uploading or pasting anything, minimize personal data. Remove or obscure your full date of birth, home address, policy number, medical-record number, payment information, and unrelated medical history. The safest useful record is usually the smallest one that contains the facts needed for the task.

3. Ask for Structure, Not a Diagnosis

A useful request might read:

Transcribe this medical-office voicemail. Then list the caller, purpose, medications mentioned, requested action, contact method, deadline, and any details that are uncertain. Do not provide medical advice or invent missing information. Draft a concise reply that asks the office to confirm the treatment and prior-authorization next steps.

4. Verify Every Important Detail

Replay the audio while reading the transcript. Check medication spelling, dosage, frequency, names, telephone numbers, email addresses, and dates. If the recording is unclear, mark the detail as uncertain instead of guessing. A polished error is still an error.

5. Make the Office’s Next Action Obvious

Keep the final message short enough to scan. State the problem, provide the relevant history you can verify, explain what the pharmacy or insurer told you, and ask what the office needs from you. Questions about prior authorization, covered alternatives, temporary samples, and follow-up testing belong with the clinician and office team.

6. Save the Record

Keep the original voicemail or letter, the verified transcript, the final message you sent, and any reply. Record the date and the next promised action. AI is helpful here because it can turn scattered communications into a simple timeline and checklist.

What AI Must Not Do

This workflow has clear boundaries. A general AI assistant should not decide that a medication is safe to stop, tell a patient to substitute one glaucoma drop for another, change a dosage, diagnose worsening disease, or promise that a prior authorization will be approved.

Glaucoma can damage vision without dramatic symptoms. Questions about continuing medication, using an alternative, receiving samples, or waiting for authorization require prompt guidance from the prescribing eye-care professional. New eye pain, sudden vision changes, halos, severe headache, nausea, or other urgent symptoms should be handled through appropriate medical or emergency care—not a chatbot or an email draft.

AI also makes mistakes. Transcription can fail when audio is muffled. A model can confidently “correct” a medication into the wrong word. It may omit a qualification that mattered to the caller. The answer is not blind trust; it is a workflow in which AI does the first pass and the patient performs the final verification.

This Is What Human-in-the-Loop Healthcare Looks Like

The phrase human in the loop can sound like something invented for a technology conference. Here, it means something very practical:

• The doctor remains responsible for medical judgment.
• The office handles the authorization and clinical documentation.
• The pharmacy reports the coverage rejection.
• The insurer or benefit manager applies the coverage rules.
• The AI organizes and drafts.
• The patient verifies, approves, sends, and follows up.

That division of responsibility is not as flashy as asking AI to “solve healthcare.” It is more believable—and useful right now.

From One Personal Problem to a Reusable AI Tool

This experience also reflects how we build tools at AI Workflow Solutions, LLC. We start with a real obstacle, observe where the information becomes confusing, and design a repeatable process that leaves the person in control.

The same pattern could support many non-diagnostic healthcare tasks: preparing questions for an appointment, comparing a bill with an explanation of benefits, organizing discharge instructions, building a medication list for review, summarizing questions for a caregiver meeting, or tracking which office promised to call next.

The goal is not to make patients practice medicine. It is to help them arrive prepared, communicate clearly, and preserve an accurate record. Used carefully, AI can reduce the administrative fog surrounding healthcare so the human conversation can focus on care.

The Real Medical-AI Revolution May Begin With Better Homework

Our podcast looked toward a future in which AI may become deeply integrated into medicine. My denied-prescription experience showed me a smaller version of that future already taking shape.

A voicemail became a transcript. The transcript became an action list. The action list became a professional email. The email gave the doctor’s office a clear request it could move forward. No diagnosis was automated, and no clinician was replaced. A frustrating communication gap simply became more manageable.

That is the kind of AI progress I trust: not a machine pretending to know everything, but a tool helping a person ask better questions, preserve the facts, and take the next responsible step.

Have you ever left a medical voicemail, insurance call, or pharmacy conversation unsure what to do next? Share this article with someone who could use a calmer way to organize the process, and subscribe to Deep Dive AI on YouTube for more practical experiments showing what today’s AI tools can—and cannot—do.

You can also follow AI Workflow Solutions on Facebook and listen to the Deep Dive AI podcast on Spotify.

---

Medical disclaimer: This article describes a personal communication and organization workflow for educational purposes. It is not medical advice and does not replace guidance from a licensed clinician, pharmacist, insurer, or emergency service. Never start, stop, substitute, or change a prescription medication based solely on AI-generated content.

AI and privacy disclosure: Use the minimum information necessary, review the privacy terms of any AI service you choose, remove unnecessary identifying information, and verify all generated transcripts and drafts before sending them.