Circle Medical built an AI scribe in two weeks with Recall.ai

Executive Summary

Challenge

Circle Medical wanted to reduce the time clinicians spent documenting patient encounters.

The team first tried using browser tools to capture audio and generate transcripts, but the approach quickly became clunky and difficult to maintain. Building reliable integrations with Zoom and other video platforms would have required a much larger engineering investment.

They needed infrastructure that could scale with their virtual care platform, support HIPAA-compliant workflows, and let them choose a transcription provider capable of handling medical terminology.

Solution

Circle Medical used Recall.ai to build an AI scribe that automatically joins virtual patient encounters and returns the recording and transcript to its systems.

The team then applies its own AI to extract clinical information and generate a first draft of the patient note directly in its EHR.

Two people built the feature, with one engineer integrating Recall.ai in approximately two weeks. This allowed them to focus on the healthcare-specific parts of the product: clinical analysis, AI-generated notes, and the EHR workflow.

Results

With Recall.ai powering its AI scribe, Circle Medical:

• Reduced the time providers spend completing clinical notes by approximately 75%
• Saved close to 50,000 hours of clinical documentation time within months of launch
• Scaled the AI scribe across millions of patient encounters
• Built the feature with two people in approximately two weeks
• Automatically added clinical note drafts to its EHR
• Maintained control over which transcription provider it used for medical terminology
• Built on HIPAA-compliant infrastructure backed by a BAA

How Circle Medical’s healthcare AI scribe works

Circle Medical is a modern primary care clinic with a network of providers across the US serving over 55,000 patients each month both in-person and virtually.

Its healthcare AI scribe, sometimes called an automated clinical notetaker, uses Recall.ai to capture virtual patient encounters and return the recording and transcript. Circle Medical then applies its own AI to extract clinical information and generate a first draft of the patient note inside its EHR.

The workflow is straightforward:

Virtual patient encounter → Recall.ai recording and transcript → Circle Medical’s AI → clinical note draft in the EHR

This allows providers to focus on the patient during the visit instead of documenting every detail as the conversation happens.

“Using Recall.ai really let us focus on the real problem that we wanted to solve,” said Miguel Jetté, Head of AI at Circle Medical. “We wanted to create a productivity tool for our providers to help them spend less time documenting their encounters.”

Browser-based audio capture was too clunky to scale

Because most Circle Medical providers see patients virtually, the team initially tried using browser tools to capture audio and generate transcripts.

The approach worked as an early test, but it was not reliable or maintainable enough for a production healthcare product.

“Before Recall.ai, we used browser tools to capture the audio when the provider would speak with their patients, trying to get a transcript out of it,” Miguel said. “It was really clunky, it was very complicated, and so we knew we needed a better solution.”

They also realized that supporting Zoom and other video platforms would become a substantial engineering project. The infrastructure needed to work across different provider setups and continue scaling as Circle Medical grew.

“We quickly realized that building the integration with Zoom and all the other video providers was going to be very complicated,” Miguel said. “We wanted a solution that scales, that’s reliable, and that’s future-proof.”

Recall.ai gave them that foundation without requiring the team to build and maintain the meeting infrastructure itself.

Two people built the healthcare AI scribe in two weeks

Circle Medical built the AI scribe with a two-person team. One engineer spent approximately two weeks integrating Recall.ai, work that Miguel estimated could otherwise have taken a team of five or six engineers several months.

“We were able to build this feature with only two of us on the team,” Miguel said. “One engineer spent maybe two weeks integrating Recall.ai, something that would have taken us months with a team of five or six people.”

Recall.ai handled getting the bot into the encounter and returning the recording and transcript. The team could spend its time on the AI, clinical workflows, and EHR integration that made the product useful to providers.

The integration itself was straightforward. Recall.ai’s documentation was easy to follow, and a shared Slack channel gave the team a direct place to ask questions and receive fast answers.

“Working with Recall.ai was very easy,” Miguel said. “Documentation is easy to follow. I have nothing but good things to say about working with Recall.”

Automatically recording virtual visits and adding patient notes to the EHR

Once a virtual encounter ends, Circle Medical processes the transcript using its own AI systems.

The AI scribe extracts information such as the history of present illness, prescribed medications, and the plan for following up with the patient. It then generates a first draft of the clinical note directly inside the EHR, where the provider can review and complete it.

“We integrated Recall.ai and that whole workflow into our EHR system,” Miguel said. “Right after the encounter, the provider will have a first draft of the whole clinical note.”

The encounter also becomes part of the patient history already stored in Circle Medical’s health system. Providers can use that context to personalize future care plans and better understand what happened across previous visits.

This was where Circle Medical wanted its engineers spending their time. The value was not in rebuilding the infrastructure needed to join and record a video call. It was in turning the visit into useful clinical information and delivering it inside the workflow providers already used.

The AI scribe reduced clinical documentation time by 75%

Circle Medical estimates that providers now spend approximately 75% less time taking notes and completing clinical documentation.

Within months of launching the AI scribe, Miguel estimated that in the first several months after launching the feature had already saved close to 50,000 hours of clinical note writing. The product has since scaled across millions of patient encounters.

Because providers do not need to document every detail during the visit, they can focus more closely on the patient. The resulting notes are also more complete and thorough, while the transcript and recording remain available for review when needed.

“The power of this tool is that now doctors can really focus on their patients,” Miguel said.

The response from providers was immediate. Some told the team they did not know how they had worked before the AI scribe existed. Receiving a draft note immediately after an encounter made documentation feel like part of the visit rather than another task waiting at the end of the day.

Medical AI scribes need transcription built for medical language

Medical encounters include drug names, conditions, procedures, abbreviations, and other terminology that general-purpose transcription systems may not recognize correctly.

Circle Medical therefore needed control over which transcription provider it used. Many meeting and notetaking products bundle a single transcription model into the service, even when that model is not designed for medical language.

Recall.ai allowed the team to choose its preferred transcription provider rather than relying on native Zoom or Google Meet transcription. Circle Medical selected Rev, but the team can continue adapting its transcription setup as its needs change.

“One nice thing about Recall.ai was that you could choose your third party for transcription,” Miguel said. “In the medical field, there’s a lot of weird terminology that maybe Zoom transcription won’t get or Google Meet transcription won’t get.”

That flexibility gives healthtech companies the option to use transcription services developed specifically for medical use cases or models that can be customized around specialized vocabulary.

Recall.ai supports HIPAA-compliant AI scribe workflows

Patient encounters contain highly sensitive health information. For Circle Medical, security and compliance were requirements from the beginning, not features that could be added after the AI scribe was built.

The team needed a provider that took data security seriously, supported HIPAA-compliant workflows, and would sign a business associate agreement.

“An important factor for us when deciding which technology to use was one that promoted high security and one that was HIPAA compliant,” Miguel said. “We deal with a lot of sensitive data every day, and we wanted a partner that takes data security seriously.”

Recall.ai met those requirements and signed a BAA with Circle Medical, allowing the team to build the AI scribe on infrastructure appropriate for its healthcare data and clinical workflows.

Circle Medical scaled its AI scribe to millions of encounters

Circle Medical needed infrastructure that could support both a fast initial launch and long-term growth.

The team has now used Recall.ai across millions of patient encounters and expects to build more features using its APIs over time. Because Recall.ai handles the meeting layer, Circle Medical can expand the product without continually rebuilding the capture infrastructure underneath it.

The same foundation can support new transcription models, deeper EHR integrations, additional clinical analysis, and new tools for providers.

“What’s amazing about Recall.ai is that it’s a versatile API that you can integrate into your product,” Miguel said. “It’s all really super easy to integrate into your own software, and it makes it really fast to build an amazing product.”

“It is by far the best productivity feature we built at Circle Medical”

Circle Medical set out to reduce the administrative burden on clinicians and give them more time to focus on patients.

With Recall.ai, a two-person team built a healthcare AI scribe in approximately two weeks, added clinical note drafts directly to its EHR, and scaled the product across millions of patient encounters. Providers now spend approximately 75% less time completing clinical documentation.

“It is by far the best productivity feature we built at Circle Medical,” Miguel said.