June 2025: Patients need an AI scribe of their own

AI scribes have raised billions to help doctors. Whose building one for patients?

A common “workflow” around a complex illness looks like:

• Patient + caregiver keeps a binder of as much medical information as possible (medical records, prescriptions, notes etc). 

• Patient + caregiver spend a lot of time researching details around the diagnosis: available therapies, known side effects, lifestyle changes, other specialists etc. 

• Patient + caregiver collaborate around care. This includes everything from logistics (who accompanies the patient to each visit) to lifestyle changes, supplements etc.

This isn’t exhaustive but it is directionally right. I think it’s now possible to combine all of these into a single tool, that the patient and caregiver will use. This would look like:

1. AI scribe on the patients phone (very similar to the Abridge patient app). Records the visit, extracts key clinical context, generates reminders and timelines.

2. Integrated Deep Research agent that 

   1. utilizes the specific visit transcript + patient demographic profile + summary + the historical visit notes as context, and makes it easy to add more context

   2. Generates a custom prompt to grab information from the outside world that the patient would want. This can vary based on the complexity of the condition, but would include everything from specialists, clinical trials, alternative therapies, research in other languages etc

   3. Enables the patient and caregivers to quickly do new research with the patients history as context.

3. Collaboration tools (this can be as simple as a WhatsApp style group chat that is integrated into the experience) to enable the patient and caregiver to just talk about the logistics and substance of care.

I suspect you could do lots more interesting things like integrate into MyChart, integrate billing etc, but ultimately these 3 pieces of functionality can be built permissionlessly and actually belong together.

I wrote a while ago about “Voice” as a new healthcare primitive, based on my experience using Abridge as a patient and on Carbon’s experience building a fast growing AI scribe. At the time the enterprise opportunity felt pretty obvious because I could directly observe the clinical and operational impact Carbon’s AI scribe had. 

The provider impact was obvious to me, but seeing the explosion in deep research agents, it’s pretty clear they can lift a ton of cognitive load for patients as well. The other nice thing is that this problem is pretty uniform for patients around the world. If you’re building this I’d love to learn more (and be an early adopter).