Giving clinicians their evenings back
How provider organizations use ambient documentation, prior-authorization automation and clinical search to reduce administrative load.
Figures: The Permanente Medical Group, 2023–2024. Source and caveats below.
01 The problem
Physicians routinely spend hours on the electronic health record for every clinic day, much of it after hours. That documentation load is a leading cause of burnout, and burnout leads to reduced hours and departures that are very costly to replace.
On the administrative side, prior authorizations, coding and denials consume staff time and delay both care and payment. The information needed is usually already in the chart, but someone has to find it and re-enter it.
02 Where AI fits
Ambient clinical documentation
With patient consent, the visit conversation is transcribed and drafted into a structured note that the clinician reviews, edits and signs in the EHR.
Prior authorization
An agent assembles the clinical evidence from the chart against the payer's criteria and drafts the request for staff to submit.
Coding and denial management
Models suggest codes with supporting documentation and draft appeal letters for denied claims.
Clinical knowledge search
Clinicians query guidelines, formularies and internal protocols and get an answer with the source cited.
Patient access
Scheduling, reminders and routine questions handled by an assistant, with handoff to staff for anything clinical.
03 Reference architecture
The layers we put in place for a typical healthcare deployment, from your systems of record up to the people who review the output.
- L1EHR via FHIR / HL7 (Epic, Oracle Health, athenahealth)
- L2Consent capture and audio transcription
- L3PHI-safe model endpoint under a BAA
- L4Note drafting and coding models
- L5Clinician review and sign-off
- L6Audit log and quality sampling
04 Real-world implementation
The Permanente Medical Group in Northern California rolled out ambient AI scribes from late 2023. In the first year, 7,260 physicians used the technology across more than 2.5 million patient encounters. A follow-up analysis published in NEJM Catalyst estimated 15,791 hours of documentation time saved, the equivalent of 1,794 eight-hour working days, and 84% of physicians reported a positive effect on communication with patients.
Source: NEJM Catalyst (2025); summarized by the American Medical Association. The Permanente Medical Group is not a 925 Techies client.
What to keep in mind: Time savings are estimates comparing users with non-users, and were greatest for the heaviest users. Drafted notes can contain errors, so clinician review before signing remains essential.
05 Business impact
Clinician time
Less time on notes during and after clinic, which is the most direct lever on burnout and retention.
Revenue cycle
More complete documentation supports accurate coding, fewer denials and faster payment.
Access
Time recovered from administration can be returned to patient care or to a sustainable workday.
Patient experience
Clinicians face the patient rather than the screen during the visit.
06 Implementation roadmap
- Weeks 1–3
Assess
Privacy and security review, BAA coverage, EHR integration path. Choose one specialty or clinic for the pilot.
- Weeks 4–9
Prove
A volunteer group of clinicians uses the tool. Measure note time, edit rate and note quality by blinded review.
- Weeks 10–20
Build
EHR integration, specialty templates, training materials and a support model for wider rollout.
- Ongoing
Operate
Sample notes for accuracy, monitor after-hours EHR time, extend to more specialties.
07 Guardrails
- HIPAA controls and a BAA with every model provider
- Patient consent before recording
- Clinician reviews and signs every note
- No autonomous diagnosis or treatment decisions
08 Services involved
AI Integration & Implementation
AI wired into the CRM, ERP, ATS and line-of-business systems your teams already use.
Small Language Models (SLMs)
Compact models fine-tuned for one job, running in your environment, at a fraction of the per-task cost.
AI Security & Governance
Guardrails, red-teaming and policy so your AI stands up to auditors, regulators and attackers.
Start here
Exploring AI in healthcare?
We can run a proof of concept on your own data in about four weeks, measured against the KPIs above.
