How AltaMed Put Ambient AI Scribes Into Daily Practice - Without Breaking Care
AltaMed serves more than 400,000 patients across 60+ sites, and like many large health centers, documentation was grinding down clinical time. Dr. Eric Lee, medical director of clinical informatics, led a rollout of ambient AI scribes that actually stuck - because the team did the unglamorous work first: research, pilots, governance, and deep EHR integration.
He's frank about the noise around AI in health IT. There are plenty of proposed use cases. Most won't matter if your data foundation is messy and your workflows aren't respected. His rule: go slow to go fast.
Why start with ambient AI scribes
Daily documentation is a real tax on providers. That's why ambient AI scribing has become the first practical win in clinical settings. It reduces the burden while preserving a private, human interaction between provider and patient.
Other popular ideas - like drafting denial or appeal letters - often pit payer AI against provider AI. That's a dead end for clinicians. Lower-risk, patient-facing use cases (scheduling, rescheduling, cancellations) can help, but only if data governance is in place. Otherwise, expect "AI slop."
Go slow to go fast: the rollout playbook
The informatics team evaluated multiple vendors and piloted features themselves. They weren't looking for hype - they wanted intuitive tools that required close to zero training and fit cleanly into existing note templates.
Deep EHR integration came next. AltaMed mapped the scribe across standard note templates and Express Lanes to keep clicks low and adoption high. Before go-live, they worked through the hard questions with leaders: Does it actually work? How are other languages handled? Where are recordings stored, and for how long? What are the ethics and consent requirements?
Governance before gadgets
AltaMed set up interdisciplinary AI governance, published AI policies, and educated providers. Access to the integrated AI scribe required a signed responsible use agreement - a simple way to set expectations from day one.
The responsible use agreement providers signed
- Always ask for consent to record.
- Use the tool only to assist with documentation; it does not provide medical advice, diagnosis, or treatment recommendations.
- Always review and edit generated documentation for accuracy.
- Do not include umbrella statements that excuse errors by blaming generative AI.
- Document that the encounter was co-created with generative AI and attest to careful review before signing.
Adoption and impact
After more than six months live, nearly three quarters of providers enrolled. Of those, over half are high utilizers, using the scribe for more than 70% of visits; another third use it 30-70% of the time.
Average time saved: about 20 minutes per day per provider. More than 95% of providers edit the generated notes - which is exactly what you want for quality and safety. The team is also tracking accuracy with other languages to protect patient experience and clinical clarity.
AltaMed is partnering with the California Health Care Foundation to study ROI so other FQHCs can replicate what works. CHCF has long supported practical innovation in community health.
Beyond scribing, AltaMed launched augmented draft responses for patient messages within its centralized nursing team, which handles roughly two-thirds of provider-directed messages. Early result: lower cognitive load for nurses and faster processing.
A simple checklist you can copy
- Pick high-friction workflows first: ambient scribe for visits, draft responses for inbound messages.
- Pilot with your informatics team before broad rollout. Test multilingual scenarios and edge cases.
- Stand up AI governance, publish clear policies, and require a responsible use agreement.
- Integrate deeply with your EHR: standard note templates, macros, and Express Lanes. Eliminate toggling.
- Data governance first. Bad data equals bad patient experience.
- Measure more than minutes: provider well-being, documentation quality, throughput, and message turnaround.
- Define recording storage and retention. Be explicit about consent and privacy.
- Make human review mandatory. No blanket disclaimers shifting blame to AI.
- Keep training light but available. Short guides, quick videos, and office hours beat long webinars.
Context that still matters
This progress sits on top of a decade of EHR work and Promoting Interoperability requirements. If your templates, note types, and data flows are clean, AI will help. If they aren't, fix that first. For background, see CMS Promoting Interoperability Programs here.
Want to level up your team's AI skills?
If you're standing up governance, evaluating vendors, or building provider training, curated courses can shorten the learning curve. Explore role-based AI tracks at Complete AI Training or browse the latest healthcare-relevant AI courses here.
The takeaway: Ambient AI scribes can work at scale in community health - with governance, consent, tight EHR integration, and relentless editing. Go slow, set guardrails, then let the workflow wins stack up.
Your membership also unlocks:
