The word integration is becoming surprisingly difficult to define in clinical AI. In actuality, when an AI scribe generates a note and lets you copy it into Med Access. That’s an integration.
When another tool sends a completed note back to the EMR without you pasting it, that is also integration.
When a tool sits alongside the patient encounter, reads relevant information already in the chart, scribes the conversation, fills the appropriate encounter fields, helps update the chart, and supports billing before you move to the next patient. That is also an integration.
Although these are three very different experiences, and three different levels of integration. But certainly, these are all integrations.
For a Canadian family physician trying to reduce administrative work, the distinction is not academic. The distinction is in how much work that is actually removed from the clinic day.
A physician can have an “integrated” AI scribe and still spend the evening sorting notes, moving between tabs, updating charts and completing billing.
That is why physicians evaluating Clinical AI integration with Med Access should look beyond whether a tool uses the word integration or not. Rather, they should consider what exactly the clinical AI does once it is connected to my Med Access workflow.
Here are three levels of AI Scribe integration you should know and consider when choosing AI Scribe for Med Access.
AI Scribe Beside Your EMR (Level 1)
This is where much of clinical AI began. The physician works in Med Access, while the AI operates somewhere else. Usually, that means a separate browser tab, web application, or device.
The workflow looks something like this:
Open patient in Med Access → open AI scribe → start encounter → generate note → review note → copy note → return to Med Access → paste note → complete remaining chart work
There is nothing inherently wrong with this model. In fact, it can still save physicians considerable typing. But it is okay to call it what it is – stressful or archaic.
The AI is beside the EMR, rather than meaningfully operating within the encounter. And that distinction starts to matter after the novelty of automatic note generation wears off.
The hidden cost of the second tab.
No individual copy-and-paste action feels particularly burdensome. Neither does switching browser tabs. The problem is repetition. Do it once, and it is trivial.
Do it after 25 patient encounters, and those small transitions become part of the administrative workload of the day. There is also a context problem.
A clinical AI tool sitting outside the EMR may know everything said during the current conversation but very little about the longitudinal patient record unless information is manually supplied to it.
For family medicine, that matters. Today’s diabetes follow-up is connected to the previous A1C. Today’s medication discussion is connected to what the patient was prescribed six months ago. Today’s blood pressure is connected to a trend.
Primary care is longitudinal by nature. A clinical AI assistant that only hears today’s conversation understands the visit. It does not necessarily understand the patient.
AI Sends Something to Your EMR (Level 2)
The second level is more sophisticated. The AI still exists largely outside the encounter workflow, but information can move back into Med Access. This is often described as write-back, push-to-EMR, EMR connectivity, or simply integration.
And it is certainly an improvement over copying and pasting everything manually. But physicians should look carefully at what is actually being written back.
Med Access is highly configurable, and encounter templates can contain multiple structured fields. A SOAP template, for example, may separate Subjective, Objective, Assessment and Plan rather than treating the clinical note as one large text box.
Suppose your AI scribe generates a perfectly structured SOAP note. Then it sends the entire note into the first available field. Technically, information has moved from the AI into Med Access. But what happens next? You may still need to split the note. You may still need to update medications or the problem list. You may still need to determine billing codes. You may still need to navigate through Med Access to complete the rest of the encounter.
This is why “Does it write to Med Access?” is no longer a sufficient question when evaluating clinical AI. What is relevant are:
What does it write?
Where does it write it?
Does it understand my encounter template?
Does it understand which patient I currently have open?
Can it use information already in the chart?
What work remains after the write-back happens?
When you ask yourself these questions, you will get the understanding of the second level of integration.
3. AI Works Within the Encounter (Level 3)
This is where the meaning of clinical AI assistant integration becomes much more interesting. Instead of treating the EMR as somewhere to send a completed AI output, the clinical AI becomes part of the encounter itself.
The patient chart provides context. The conversation provides new information. The AI scribe helps the physician turn both into different pieces of work required to complete the visit.
The workflow becomes closer to:
Open patient encounter → AI works alongside the chart → conduct visit → review documentation and suggested actions → complete encounter
The physician remains in control. But the AI Scribe is no longer another destination in the workflow. It becomes part of the workflow. That is a fundamentally different model.
What Should a Deeply Integrated AI Scribe Actually Do?
There is no single feature that makes a clinical AI assistant tool deeply integrated. It is the relationship between the AI, the patient chart, and the encounter workflow. For a family physician, there are several useful tests to confirm your clinical AI integration is on level 3:
1. It should know which encounter you are working on
This sounds obvious. It is not. An encounter-level integration should be connected to the patient chart the physician already has open. That removes the need to repeatedly identify, select or match the patient across different systems.
Medulla, for example, opens as a drawer alongside the Med Access patient encounter. The scribe and generated documentation are therefore associated with the encounter already open rather than being created in a separate workspace that must later be reconciled.
That is what complete integration should feel like. Not “I have connected two pieces of software.”, but “I am still working on the same patient.”
2. It should understand the structure of Med Access
Med Access templates exist for a reason. A clinic may have spent years adapting them to the way its physicians document care. Good clinical AI integration should respect that structure rather than forcing physicians to redesign their workflow around the clinical AI.
If an encounter uses separate Subjective, Objective, Assessment and Plan fields, for example, the clinical AI should ideally understand those destinations.
Medulla’s current Med Access workflow splits generated documentation across the corresponding fields rather than placing the entire note into one field.
This sounds like a small technical distinction. Operationally, it is very significant.
A physician should not have to destroy the structure of a carefully designed EMR template simply to accommodate an AI scribe. The clinical AI should adapt to the physician’s workflow, not the other way around.
3. Integration Should Be Two-Way
This may be the most important distinction between basic write-back and deeper clinical AI integration. Many integrations focus on one direction:
AI → EMR
The AI creates something. The EMR receives it.
But useful clinical AI increasingly needs information flowing in the other direction:
EMR → AI
That changes what the AI can understand.
Instead of only knowing what was said during today’s appointment, it can work with relevant information already contained in the patient’s record.
For example, Medulla’s AI Insight can synthesize information from the existing chart, including previous encounters, medications, labs and problems, and surface potential priorities when the encounter is opened.
This matters particularly in longitudinal family medicine. A physician rarely sees a patient as a blank page. The chart is part of the clinical context. True encounter-level integration should recognize that.
4. The AI Should Help With More Than the Note
The clinical note is important. It is also only one piece of completing an encounter. After the conversation, a family physician may still need to: update the problem list, record medication changes, add follow-up plans, review clinical information, complete billing, and, depending on the encounter, perform other administrative tasks.
If an AI scribe produces a note but leaves every surrounding task untouched, it has automated transcription more than it has integrated with the clinical workflow. This is where the distinction between an AI scribe and an AI clinical assistant starts becoming meaningful.
Medulla’s current Med Access workflow combines several encounter-level functions. It can generate the note, propose chart updates, provide chart-aware clinical answers, suggest billing and diagnostic codes and support telephone visits from the encounter workflow.
The physician still reviews and decides what belongs in the medical record. That human control is important. The objective of integration should not be autonomous medicine. It should be less or zero manual administration around physician-led medicine.
5. Billing Should Be Connected to What Actually Happened
Billing is often treated as something that happens after documentation. In practice, both originate from the same encounter. The conversation establishes what happened. The note documents it. The billing reflects the work performed.
When those workflows are disconnected, the physician may find themselves reconstructing the encounter later in order to choose the appropriate diagnostic or service codes. An encounter-integrated AI can reduce that separation.
Medulla, for example, surfaces diagnostic and billing-code suggestions based on the encounter and provincial fee schedule as the physician finishes the visit.
The physician reviews the suggestions before applying them.
Again, the value is not simply that “AI can do billing.” The value is that another task can potentially be completed while the encounter is still active, rather than becoming end-of-day work.
6. Phone Visits Should Still Feel Like the Same Workflow
Telephone visits provide a useful test of how integrated a clinical AI platform really is.
Without integration, the workflow can fragment quickly:
Open the chart.
Find the patient’s phone number.
Call.
Open the AI scribe.
Document.
Return to Med Access.
Complete the encounter.
With encounter-level integration, the phone visit can remain connected to the chart.
Medulla’s Med Access workflow can pull the patient’s phone number from the chart, place the call using the clinic’s caller ID and scribe the telephone encounter while it happens.
Again, the individual feature is less important than the principle behind it:
Information that already exists in the EMR should not need to be manually entered into another tool just to complete the same patient encounter.
7 Questions to Ask Before Choosing an AI Integration for Med Access
If you are comparing AI scribes or clinical assistants for your Med Access practice, skip the marketing terminology and ask for a live demonstration.
Then ask these questions:
- Does the AI work on my Med Access encounter page, or in a separate application?
- Can it read relevant context from the patient chart, or only what I manually give it?
- Does it write into the correct Med Access template fields?
- What do I still need to copy, paste or re-enter?
- Can it help update the chart beyond generating the encounter note?
- Can it support billing and phone encounters within the same workflow?
- What work will still be waiting for me after my last patient?
That seventh question may be the most revealing. Because integration should ultimately be judged by its effect on the physician’s day.
Why This Matters for Canadian Family Physicians
Family medicine is unusually sensitive to workflow friction. A family physician may move through dozens of encounters in a day, while simultaneously managing longitudinal histories, medications, preventive care, chronic disease, results, referrals and follow-up.
A workflow that adds only a few unnecessary actions to one visit can multiply those actions across an entire clinic. This is why the next generation of clinical AI should not simply be measured by how impressive its note generation is.
The better measure is: How much of the encounter can the physician finish before moving to the next patient?
That is a much higher standard. And it is one worth demanding.
Medulla’s Approach to Med Access AI Integration
Medulla was designed around what we would call Level 3 integration. It does not ask the physician to leave Med Access and conduct the visit inside another application.
Instead, Medulla opens alongside the patient chart on the Med Access encounter page.
From there, the workflow can include:
AI Insight to synthesize relevant information already in the patient’s chart.
AI Scribe to capture the encounter and generate a concise clinical note.
Structured template filling so Subjective, Objective, Assessment and Plan can land in their corresponding Med Access fields.
AI Autochart to propose relevant updates to problems, medications, care plans and follow-ups for physician approval.
AI Billing Assistant to suggest diagnostic and service codes based on what occurred during the encounter.
Ask to answer clinical questions using patient context and cited Canadian sources.
Telehealth to call patients from the encounter workflow while keeping the visit connected to the chart.
The important part is not the length of that feature list. It is that these functions revolve around the same patient encounter. That is the difference between adding clinical AI to an existing workflow and designing clinical AI around the workflow itself. See Medulla inside Med Access
Charting done before the patient leaves the room.
Medulla works inside Med Access and Oscar-based EMRs, on the encounter you already have open.
No credit card required.