The next phase of healthcare AI is not about opening ChatGPT in another browser tab. It is about AI gaining access to the clinical context where healthcare decisions are made.
On September 1, OpenAI announced something that pharmacists should pay very close attention to.
Healthcare organizations can now connect authorized patient information from their Epic electronic health record directly with ChatGPT for Healthcare.
In supported deployments, ChatGPT can also be integrated directly into the EHR layout, allowing AI-assisted workflows without leaving the patient chart.
At the same time, OpenAI launched a Healthcare Public Data plugin that connects ChatGPT and Codex with nine official healthcare data sources:
PubMed
ClinicalTrials.gov
DailyMed
RxNorm
openFDA
CMS Coverage
CMS Open Data
Medicare Care Compare
NPI Registry
And buried inside OpenAI’s announcement is an example that should immediately catch the attention of pharmacists.
OpenAI specifically describes a pharmacy team using DailyMed to confirm the latest medication label and warnings.
That may sound like a small example.
It isn’t.
Because this represents a much larger shift in what AI could look like inside pharmacy practice.
AI is moving from the browser window into the workflow
Until recently, much of the conversation around generative AI in pharmacy looked something like this:
Open ChatGPT.
Type a question.
Ask it to summarize something.
Draft a document.
Research a topic.
Explain a guideline.
Then take that information back into whatever system you were actually working inside.
Useful?
Absolutely.
Transformative?
Potentially.
But still separate from the actual clinical workflow.
The September 1 announcement moves us into a different phase.
An authorized user in a properly configured healthcare organization can now use ChatGPT to review Epic information they already have permission to access, including clinical notes, medications, conditions, encounters, and laboratory results.
OpenAI gives examples such as asking:
What changed since the patient’s previous visit?
Which recent laboratory results should be reviewed?
Have there been medication changes?
Are there new specialist recommendations?
What follow-ups or unresolved issues remain?
ChatGPT can bring together information from the authorized patient record, summarize relevant developments, and link back to the supporting chart information.
Now think about what that means from a pharmacy perspective.
Medication list.
Laboratory results.
Conditions.
Clinical notes.
Recent encounters.
Specialist documentation.
Official medication labeling.
Published literature.
Clinical trials.
Medicare coverage information.
We are getting closer to AI systems that can operate where the information pharmacists spend enormous amounts of time searching, organizing, reviewing, and reconciling already lives.
That is a fundamentally different use of AI.
The pharmacist’s opportunity moves higher up the stack
Imagine a pharmacist conducting a medication review today.
A substantial amount of time may be spent gathering information.
Finding the medication history.
Reviewing recent laboratory values.
Finding the last specialist note.
Checking whether a medication was discontinued.
Searching the package insert.
Looking up renal dosing.
Reviewing recent literature.
Checking coverage requirements.
Building a clinical picture from information scattered across multiple systems.
AI can potentially make portions of that information retrieval and synthesis dramatically faster.
That does not mean the pharmacist becomes less important.
It means the valuable part of the pharmacist’s work begins to shift.
If information becomes easier to retrieve, the differentiating skill becomes knowing:
Is this information correct?
Does it matter for this patient?
What is missing?
What conflicts with something else in the chart?
What should actually be acted on?
What risks does the AI not appreciate?
When should the recommendation be challenged?
What requires communication with the physician?
What should absolutely remain a human decision?
That is a very different skill set from simply being good at finding information.
And pharmacists are unusually well positioned for it.
AI does not eliminate the medication expert. It changes which part of medication expertise becomes most valuable.
This distinction matters.
For decades, part of being highly knowledgeable in pharmacy meant knowing where to find the answer quickly.
Which reference?
Which guideline?
Which package insert?
Which interaction database?
Which formulary?
Which section of the chart?
Those skills are still useful.
But when AI can increasingly retrieve, organize, and summarize that information, the relative value of information retrieval alone is likely to decline.
What becomes more valuable?
Validation
Can you determine whether the output is clinically accurate?
Judgment
Can you decide whether the information actually applies to this patient?
Exception handling
Can you recognize when the patient in front of you does not fit the standard pathway?
Medication safety
Can you identify the risk created by the combination of medications, laboratory findings, diagnoses, age, renal function, adherence, and other clinical factors?
Workflow design
Can you determine where AI should assist and where a pharmacist must remain involved?
Governance
Can you help establish what an AI system should be allowed to do, what it should not do, and what requires human review?
These are not peripheral skills.
They may become some of the most valuable pharmacist skills in an AI-enabled healthcare system.
Even OpenAI’s documentation makes the human role clear
There is an important detail in the release that should not be ignored.
The Epic integration is currently read-only.
It cannot update the medical record.
It cannot place orders.
It cannot message patients.
It cannot override the user’s existing Epic permissions.
And OpenAI explicitly states that clinicians remain responsible for reviewing the underlying record and making care decisions.
The public healthcare sources have limitations too.
For example, OpenAI’s documentation specifically notes that medication information from DailyMed still requires clinical interpretation.
RxNorm can standardize medication names and identifiers, but it does not determine appropriate dosing, interchangeability, or formulary coverage.
OpenFDA can provide safety and regulatory information, but adverse-event data does not automatically prove that a drug caused an event.
CMS Coverage can provide Medicare coverage policies, but that does not determine an individual patient’s benefits.
That is exactly the point.
AI can bring information closer.
It can organize it.
It can summarize it.
It can help surface what might matter.
But healthcare information still needs interpretation.
And medication information requires a medication expert.
Think about medication reconciliation
Take one of the most common pharmacy workflows.
Medication reconciliation.
Today, a pharmacist may have to compare:
The medication list.
Recent hospital orders.
Outpatient prescriptions.
Specialist notes.
Discharge documentation.
Patient-reported medications.
Duplicate therapies.
Discontinued medications.
Dose changes.
Laboratory values.
The system may contain conflicting information.
AI could potentially help organize that information and highlight what changed.
But then the hard questions begin.
Was the medication truly discontinued?
Was it held temporarily?
Did the specialist intentionally change the dose?
Is the patient actually taking it?
Does the laboratory value make the current dose inappropriate?
Is the duplication intentional?
Was the medication prescribed by someone who did not know about another therapy?
Is the patient’s reported regimen more current than the chart?
Those are not search problems.
Those are judgment problems.
And this is where pharmacist expertise becomes even more important.
Or consider a renal dosing review
AI may be able to identify:
Current medications.
Recent creatinine values.
Documented kidney disease.
Medication labeling.
Relevant dosing recommendations.
Recent notes.
Potentially even recent research.
That can eliminate enormous amounts of searching.
But someone still needs to ask:
Which renal function estimate is appropriate here?
Is kidney function stable?
Is this acute kidney injury?
Is the laboratory value outdated?
Does the indication change the recommended dose?
Is the patient experiencing toxicity?
Would changing therapy create another risk?
Is the recommendation clinically meaningful enough to intervene?
That is not simply information retrieval.
That is pharmacy.
And this creates an entirely new career surface area for PharmDs
This may be one of the most exciting parts of the story.
When healthcare AI moves deeper into clinical workflows, organizations need people who understand both the clinical domain and the workflow being automated.
For medication-related workflows, pharmacists bring highly relevant expertise.
That could increasingly create opportunities around areas such as:
Clinical AI validation
Testing whether medication-related AI outputs are accurate, clinically meaningful, complete, and safe.
Medication AI governance
Helping organizations establish standards for where AI can be used, what requires pharmacist review, how errors are handled, and which medication decisions should never be automated without appropriate oversight.
Clinical informatics
Designing how medication, laboratory, diagnosis, and clinical data move through healthcare systems and into decision-support workflows.
Healthcare product development
Working with engineers, designers, clinicians, and product teams to create technology that actually works inside pharmacy and healthcare environments.
Workflow design
Identifying repetitive work that AI can assist with while preserving the points where pharmacist judgment creates the most value.
Clinical implementation
Helping health systems introduce AI tools into real clinical environments, evaluate performance, educate teams, and monitor whether the technology is improving care.
AI evaluation and safety
Developing test cases, identifying failure modes, reviewing outputs, and determining where medication-related AI systems may create risk.
Clinical operations
Building scalable systems around medication review, prior authorization, discharge, formulary management, population health, adherence, medication safety, and other pharmacy workflows.
Not every one of these jobs will require a PharmD.
But a pharmacist who combines medication expertise with informatics, AI literacy, product thinking, data, workflow design, or implementation knowledge could become extremely valuable.
Because healthcare technology companies have a recurring problem:
Technology teams understand technology.
Clinical teams understand healthcare.
Someone has to understand both well enough to translate between them.
That is an opportunity.
The skill pharmacists should start developing is not “prompt engineering”
There has been a lot of emphasis on learning how to write better AI prompts.
That can be useful.
But I think pharmacists should think bigger.
The more valuable skill may be learning how to deconstruct a clinical workflow.
Take any process inside your pharmacy and ask:
What information comes in?
What work is repetitive?
Where does judgment occur?
Where are errors most dangerous?
What information is missing?
What does the pharmacist actually decide?
What output needs to be created?
What absolutely must be verified by a human?
That is how you begin thinking like someone who can help design healthcare AI.
And you do not need to be a software engineer to start.
Your Pharmacy Unlocked challenge this week
Choose one frustrating medication-related workflow in your current job.
It could be:
Medication reconciliation.
Chart review.
Prior authorization.
Discharge medication review.
Formulary research.
Clinical intervention documentation.
Renal dosing review.
Medication history.
Patient counseling preparation.
Drug information research.
Adherence outreach.
Then map it using five boxes.
Inputs → Repetitive Work → Judgment Required → Final Output → Pharmacist Verification
1. Inputs
What information do you need?
Medication list?
Labs?
Diagnoses?
Notes?
Insurance information?
Drug labeling?
Guidelines?
Patient-reported information?
2. Repetitive work
What are you doing over and over?
Searching?
Copying?
Comparing?
Summarizing?
Sorting?
Looking through notes?
Finding laboratory values?
Checking criteria?
3. Judgment required
Where do you actually have to think?
What information must be interpreted?
What conflicts need to be resolved?
What requires clinical knowledge?
Where could the wrong decision harm the patient?
4. Final output
What are you ultimately trying to produce?
A recommendation?
A medication list?
A note?
A prior authorization?
A counseling plan?
A message to the physician?
A documented intervention?
5. Pharmacist verification
Finally, ask the most important question:
What part of this workflow should never leave the pharmacist’s responsibility?
That is where your future value may be hiding.
Do not compete with AI at the part AI is getting better at
This is the mindset shift I want pharmacists to understand.
If AI becomes excellent at finding a medication label, pharmacists should not build their professional value around being faster at finding the label.
If AI becomes excellent at summarizing twenty pages of clinical notes, pharmacists should not compete on who can summarize twenty pages faster.
Move higher.
Can you identify the clinically important detail the summary missed?
Can you recognize that the recommendation does not fit the patient?
Can you identify the medication risk hidden across three different parts of the chart?
Can you design a workflow where AI saves twenty minutes without compromising safety?
Can you determine which AI outputs require mandatory pharmacist review?
Can you validate whether a medication model is performing correctly?
Can you translate a real pharmacy workflow into something a product and engineering team can build?
Those are incredibly valuable skills.
Pharmacy has an opportunity to help design the AI layer of healthcare
There is a risk whenever technology changes quickly.
People assume their profession either wins or loses.
Reality is usually much more interesting.
Some tasks disappear.
Some tasks become easier.
Some skills become less valuable.
Others become dramatically more valuable.
And entirely new jobs emerge.
The September 1 announcement does not mean every pharmacist will suddenly have ChatGPT inside Epic tomorrow.
Access requires organizational approval, appropriate workspace configuration, Epic integration, individual authentication, existing patient permissions, and appropriate privacy and compliance controls.
But the direction is important.
AI is moving closer to the clinical record.
Closer to the medication list.
Closer to laboratory values.
Closer to official medication information.
Closer to the workflow.
And when technology moves into the workflow, healthcare organizations need people who understand that workflow deeply enough to make sure the technology actually works.
For medications, pharmacists already possess much of that domain knowledge.
The opportunity now is to layer something on top of it.
Informatics.
AI evaluation.
Product thinking.
Data.
Implementation.
Workflow design.
Governance.
Technology.
The pharmacist of the future does not have to compete with AI for who can retrieve information fastest.
The pharmacist can be the person who determines what the information means, whether the AI got it right, what should happen next, and how the system should be designed in the first place.
That is a much bigger opportunity than simply learning how to use ChatGPT.
And we are only beginning to see what it could look like.
Resources & Citations
1. OpenAI. “Healthcare organizations can now connect EHR and additional industry data to ChatGPT.” September 1, 2026. Primary announcement of the Epic EHR integration and Healthcare Public Data plugin. The announcement describes authorized Epic patient context, supported in-EHR workflows, examples involving medication changes, laboratory results and specialist recommendations, the nine-source public data capability, and the example of a pharmacy team using DailyMed to confirm medication labeling and warnings. OpenAI healthcare EHR and public data announcement 2. OpenAI Help Center. “Using the Epic plugin with ChatGPT and Codex.” Updated September 2026. Details the Epic integration’s availability, read-only status, patient information that can be reviewed, FHIR-based organizational configuration, authentication requirements, existing patient-chart permissions, privacy requirements, and the continued responsibility of clinicians to review the underlying record and make care decisions. OpenAI Epic plugin documentation 3. OpenAI Help Center. “Using Healthcare Public Data in ChatGPT and Codex.” Updated September 2026. Lists all nine official public healthcare sources: PubMed, ClinicalTrials.gov, DailyMed, RxNorm, openFDA, CMS Coverage, CMS Open Data, Medicare Care Compare, and NPI Registry. It also describes the purpose and limitations of each source and notes that medication information still requires clinical interpretation. OpenAI Healthcare Public Data documentation4. OpenAI. ChatGPT Enterprise & Edu Release Notes. “Healthcare plugins for ChatGPT and Codex.” September 1, 2026. Confirms the September 1 release of the Healthcare Public Data and Epic plugins for eligible ChatGPT for Healthcare and HIPAA-enabled ChatGPT Enterprise workspaces. The release notes also specify that Epic access is read-only, requires an administrator-configured EHR app and individual Epic authentication, and follows existing patient-chart permissions.
OpenAI release notes