Core insight
Ambient AI scribes have become standard in healthcare, and the leading tools now perform at a similar level. Success comes down to integrating the scribe with the systems a clinic already runs — the EHR, scheduling, and patient portal — and keeping its quality stable after go-live. That integration work is where a rollout succeeds or stalls.
A clinic wraps up a busy Tuesday morning. The new AI scribe tool did its job all the way through: it listened to each visit, drafted clean notes, and saved the doctors a lot of manual typing. And then… a coordinator spends the afternoon copying those notes into the medical record by hand, because the two systems were never introduced. It’s tiresome and frustrating. In a few days, half the team had stopped opening the tool.
Was the fault in the AI scribe? Should the clinic have picked another brand instead? Or maybe the onboarding process was at fault here? Three times no - the scribe worked. It just had nowhere to send the note, because nobody connected it to the systems the clinic already uses.
Ambient AI scribes are now standard
Clinicians lose a large share of their day to documentation. Tools that sit in on a visit and write the note for the doctor have moved from novelty to expectation in a couple of years.
The scale, by the providers' own numbers, is already large. One ambient scribe is reaching more than 200,000 NHS staff through a single partner. Another runs across 130-plus health systems. A vendor-commissioned study in a Swedish health system found documentation time per note dropping 29% - from 6.7 to 4.7 minutes - across a deployment of more than 375,000 notes. Most buyers no longer ask whether to adopt one, only when, and how.
Not every tool qualifies
Most tools now write a note to a similar standard, but only once they clear three requirements first: compliance and medical terminology, data security, and structured output.
The tool has to know medical language well enough to use it correctly in the record, recognizing an ICD (International Classification of Diseases) code where one applies. It also has to filter out the parts of a visit that aren't clinical - the small talk about a patient's weekend or their kids, which builds the relationship but doesn't belong in a note. Getting that split wrong either buries the record in noise or drops something the billing team needed.
Security narrows the field further. A tool has to meet the requirements of the specific hospital, clinic, or country it runs in: where data can be stored, what encryption is required, whether every access leaves a change log, and where the servers sit. Those requirements vary by jurisdiction and by institution, and a tool cleared for one clinic isn't automatically cleared for another.
Structure decides what happens to the note afterward. The more ontologies a tool builds into its summaries, the more of that content a clinic can extract as structured data automatically - for billing, quality reporting, or any other downstream system. A note without that structure is only readable by a human.
Run a shortlist through those three requirements, and what's left behind does perform at a similar level. The real test comes next. The tool that sits to the side of the daily workflow becomes one more login, one more window, one more place a busy doctor has to remember to check. Adoption erodes, and within a few weeks the clinic is paying for a tool nobody uses.
"AI scribes only work in practice when they remove work from clinicians, not add more steps. If the note is automatically saved in the right place in the medical record, clinicians are likely to keep using it. If they have to copy, paste, or manually file anything, adoption quickly drops. That is why integration is critical." - Joanna Kasprzak, PhD - COO at Apzumi
The friction is fixable, though. It stops being one more window the moment it plugs into the systems a clinician already uses - and that connection is something you can engineer.
The next step: AI scribe integration with your EHR
For a scribe to stick, it has to reach into the systems a clinic already depends on. Three connections matter most.
- The note has to land in the medical record (the EHR) on its own, formatted and filed where the next clinician expects it, so nobody has to type it in again.
- The schedule has to feed the scribe, so it knows who is in the room and attaches the note to the right patient and the right visit.
- The patient portal has to receive the after-visit summary, so the patient sees clear instructions without a staff member copying them across.
None of that arrives in the box. Each connection has to be mapped to how a specific clinic actually books, sees, documents, and follows up with patients. That mapping is the work that decides whether the scribe survives past the pilot.
Where AI scribe ROI is won or lost
Most of the buying effort goes into picking the tool. Teams sit through demos, compare note quality, and line up the pricing. People focus on how easy the tool is to use, how accurate its outputs are, and how much they can customize it. At the end, everyone is satisfied.
The return, however, comes much later, from the implementation - and that calls for a different skill than choosing software did. Selecting a tool rewards clinical judgment, procurement sense, and the ability to verify a vendor's compliance and security claims. Wiring it into the EHR, the schedule, and the patient portal rewards people who know how those systems behave. The people who ran the selection rarely have that second skill, and few clinics have it in-house.
So the decision that looks finished at signing is only half made. A scribe that nobody connects goes unused within weeks, and the time savings from the slide deck never reach the clinic. Turning that purchase into daily use is implementation work - and that's where a partner comes in.
How Apzumi helps integrate AI scribing tools
We build the connections between the tools a healthcare organization already runs and the new ones it wants to adopt. With ambient scribe, that means wiring the chosen tool into the EHR, the scheduling layer, and the patient portal, then shaping it around the way a given clinic actually works. Integration also continues after go-live: we monitor how accurately the tool performs in real use and adjust when a vendor updates the underlying model, so quality holds as the technology changes. We do not sell a scribe of our own, so our advice stays honest: pick the tool that suits you, and we'll make it work with the rest of your stack.
An AI scribe integration we built
We integrated an ambient scribe into the platform we built for LMCare, a Dutch online psychology clinic. The challenge was a fragmented stack - an EHR system, separate HR software, and a standalone video tool, none of them connected - with the scribe sitting off to the side. We wired the scribe into all of them, so after each session it drafts a formatted, GDPR-compliant report and files it where the workflow expects it, with no copying between systems. As a result of the project, admin time dropped by more than half.
The tools are ready and the demand is settled. Whether a scribe sticks comes down to one thing: connecting it to the way your clinic already works. If a rollout has stalled, or you want one to land cleanly the first time, that's the conversation we're built for. Let's talk!
FAQ
Are all ambient AI scribes basically the same?
No, there are real differences: in accuracy, and in how well a tool handles a doctor's specialty and accent. They also differ in their ability to recognize medical terminology and coding, like an ICD (International Classification of Diseases) code, and to filter non-clinical talk out of a note. Security and compliance vary too: not every tool meets a given hospital's or country's rules for data storage, encryption, or server location. Structure varies as well - it decides how much data a clinic can pull back out of a summary for billing or reporting. Among tools that meet those requirements, writing a usable note is where they perform about the same. Even the strongest one underdelivers if it isn't connected to the systems your clinic runs on.
Why do ambient scribe rollouts fail?
Usually because the tool doesn't fit the daily workflow - when the note has to be copied into the record by hand, or the app sits outside the clinician's normal flow, usage drops within weeks. Other common causes: too little training, a mismatch with certain visit types or specialties, notes that need so much editing they cost more time than they save, and weaker transcription for some accents or speaking styles.
What does a scribe need to connect to?
Three systems matter most: the electronic medical record, where the note belongs; the scheduling system, which tells the scribe who the visit is for; and the patient portal, which delivers the after-visit summary to the patient.
Is bringing in an AI scribe a clinical or an IT decision?
Both. Clinicians judge whether the notes are usable, and IT weighs security, compliance with local data rules, and how it fits your systems - few organizations choose a tool without both in the room. The part that decides whether it sticks is connecting it to your EHR, scheduling, and patient portal, which is specialist work most in-house IT teams don't take on alone. That's usually where an outside integration partner comes in.
Should we build our own scribe?
For most commercial providers there's little reason to. The market is mature and the leading tools are capable, so building from scratch means paying to recreate something you can buy today - then carrying the ongoing cost of keeping it accurate, compliant, and current as the models change. The budget goes further spent making a proven tool work inside your environment.
