How to Use AI for Medical Documentation: Clinical Notes, Scribing, and EHR Efficiency

Physicians spend an average of 15.6 hours per week on documentation and administrative tasks — nearly two full working days. For every 8 hours of scheduled patient time, physicians spend more than 5 hours in the EHR. Documentation is now the top cause of burnout for 16% of all providers and affects 26% of primary care physicians. AI ambient scribing directly attacks this problem — at a cost that starts at free and a time-to-value measured in days, not months. Here's what it actually looks like in practice.

The Documentation Burden in Numbers

Before recommending any tool, let me establish why this is worth your attention.

A physician seeing 20 patients per day and spending 15 minutes per patient on note completion is generating 5 hours of documentation daily. If even half of that happens after clinic hours — which a JAMA Network Open study confirms is the reality for most primary care physicians — that's 2.5 hours per evening, five evenings per week, that belong to paperwork instead of anything else.

The clinical consequences compound beyond the personal ones. Documentation completed under time pressure is more likely to contain errors, omissions, and template-cloned content that doesn't reflect the actual encounter. Billing codes pulled from incomplete notes undercode the visit. Prior authorization letters written from imperfect documentation require revision.

AI ambient scribing addresses the root cause rather than the symptoms. It doesn't make documentation faster — it makes documentation happen during the encounter rather than after it.

What Ambient AI Scribing Actually Does

There's a meaningful distinction between dictation and ambient scribing that determines which tool fits your workflow.

Traditional dictation: You narrate the note using structured phrasing while you work. You change how you speak and structure your thoughts. The tool transcribes what you say. Output quality is proportional to your dictation skill and discipline.

Ambient scribing: The tool listens to your natural conversation with the patient and creates the note automatically. You do not change how you speak, examine, or interact with the patient at all. The AI understands medical context, extracts the relevant clinical information, and generates a structured note in the appropriate format for your specialty.

The clinical evidence for ambient scribing is now substantial. A UCLA randomized clinical trial published in NEJM AI in November 2025, covering 238 physicians across 14 specialties, compared Microsoft DAX and Nabla against a control group. Among approximately 72,000 encounters, Nabla users saw average note-writing time drop from 4 minutes 30 seconds to 3 minutes 49 seconds — a roughly 10% reduction. Both tools showed improvements in burnout and cognitive workload.

A broader study published in JAMA Network Open in August 2025 tracked more than 1,400 clinicians at Mass General Brigham and Emory Healthcare. The burnout rate fell from 52.6% to 30.7% over 84 days at MGB. A Yale-led study confirmed that ambient AI scribes reduce physician burnout by 74% within 30 days.

These aren't marginal improvements. They're the kind of changes that determine whether a provider stays in practice.

The Pricing Landscape: What Small Practices Actually Pay

The ambient scribing market in 2026 has a wider pricing range than almost any other healthcare AI category — from free tiers to $600+/month enterprise contracts. Here's the verified 2026 landscape:

The critical pricing note for small practices: Enterprise solutions — Nuance DAX and Abridge — require minimum provider counts, typically 10–50+, and multi-year commitments, making them impractical for small practices. A 3-provider family medicine practice getting a Nuance DAX quote will be quoted $444–$600/month per provider on an annual contract — $1,300–$1,800/month for the practice. The mid-market and budget options deliver 80–90% of the same value at 20–30% of the cost.

The Right Tool by Practice Profile

Freed AI ($99–$149/month) — Best for Budget-Conscious Small Practices

Freed is purpose-built for individual practitioners and small practices who want fast setup, transparent pricing, and a simple workflow without enterprise overhead.

How it works:

• Start recording at the beginning of the encounter with one tap
• Freed captures the conversation in real time
• After the encounter, Freed generates a structured note in your preferred format
• Push the note to any browser-based EHR with one click

What it does well:

• Fastest setup of any tool in this comparison — operational in under an hour
• No IT involvement required
• Transparent month-to-month pricing with no annual contract minimum
• Learns your documentation style over time, adapting to your preferred phrasing

The honest limitation: Freed's EHR integration depth is lighter than enterprise tools. Its "EHR push" feature works via browser extension for browser-based EHRs — efficient, but not the native embedded integration that DAX provides in Epic. For practices on legacy or non-standard EHRs, this typically means copy-paste into specific fields.

HIPAA compliance: BAA available. HIPAA-compliant infrastructure.

Best for: Solo practitioners, small practices on any EHR, and providers who want to start with ambient scribing this week without an IT project.

Nabla (~$119/month paid tier) — Best Balance of Quality and Accessibility

Nabla is a European-founded AI medical scribe that has built significant clinical credibility — it was one of the two tools tested in the UCLA 2025 randomized clinical trial published in NEJM AI.

What it does:

• Ambient scribing across 55+ specialties in 35+ languages
• Works consistently across web, iOS, and Android
• Does not store audio and does not train its models on patient data by default
• Multilingual support — the strongest in the mid-market tier

What it does well:

• Privacy posture is the strongest in this comparison: HIPAA, GDPR, SOC 2 Type II, and ISO 27001 certified — the only mid-market tool with all four frameworks
• No audio storage by default — addresses the patient privacy concern that ambient scribing often raises
• Cross-device performance — works equally well on a tablet during the encounter or a desktop for review

The honest limitation: Fewer native EHR integrations than enterprise tools — copy-paste is still common in some EHR environments. Less specialty customization at the high end compared to DeepScribe or Suki.

Best for: Solo and small practice clinicians who want clinical-grade ambient scribing without enterprise pricing, multilingual practices, and any provider for whom the no-audio-storage privacy posture matters to patients.

Suki AI ($299–$399/month) — Voice Commands Beyond Documentation

Suki positions itself beyond ambient documentation to include voice commands for retrieving patient information, creating orders, generating referral letters, and ICD-10/HCC code suggestions — clinical workflow automation beyond the note itself.

What it does:

• Ambient scribing during the encounter
• Voice-activated EHR commands: "Show me the last lab results for this patient," "Add a prescription for amoxicillin 500mg"
• ICD-10 and HCC code suggestions from the generated note
• Referral letter generation from encounter summary

What it does well:

• Suki goes beyond scribing — it's a voice interface for clinical workflow that happens to generate notes. For providers who want to reduce all EHR interaction, not just documentation, this is the differentiating feature.
• Fast time-to-value: KLAS respondents report early positive results and time savings, with strongest outcomes in primary care.

The honest limitation: Specialty performance varies by site and EHR configuration. KLAS respondents ask for deeper specialty-specific capture and personalization. The $299–$399/month price point is the highest in the mid-market tier — verify the value against your specific use case before committing.

HIPAA compliance: SOC 2 Type II certified, HIPAA compliant. BAA available.

Best for: Primary care and internal medicine providers who want voice-activated EHR interaction beyond note generation — not just faster charting, but faster everything in the EHR.

Nuance DAX Copilot ($444–$600+/month) — Enterprise Standard, Not for Small Practices

For completeness: DAX Copilot is the enterprise standard, with the deepest Epic integration available. Notes appear automatically in the correct EHR fields without any provider action. This is genuinely differentiated from tools that require copy-paste or browser extension push.

However, DAX is built for hospital systems. It requires enterprise IT for setup, typically takes 2–6 weeks for full EHR configuration, and is priced for organizations where the per-provider cost is a rounding error in the technology budget.

For small practices: The ROI calculation doesn't close at $444–$600/month per provider. A 2-provider practice at $900–$1,200/month for DAX versus $200–$300/month for Freed or Nabla — with comparable documentation quality outcomes in most outpatient settings — makes DAX the wrong choice at this scale.

The one exception: if your practice is on Epic and your specific EHR configuration makes the native embedded integration meaningfully better than the copy-paste alternatives, request a demo with your specific Epic build before ruling it out.

The HIPAA Considerations Specific to Ambient Scribing

Ambient scribing introduces a specific HIPAA consideration that standard documentation tools don't: audio of the patient encounter is being captured by a third-party AI system.

Before deploying any ambient scribing tool, address these requirements:

Patient consent: Patients must be informed that an AI tool is assisting with documentation. In most states, verbal consent at the beginning of the encounter is sufficient, but some states require written consent for audio recording. Know your state's requirements before the first encounter.

Audio storage policy: Some tools store the audio recording. Others (Nabla by default, some Freed configurations) do not — only the generated text is retained. For patient privacy, no-audio-storage is the stronger posture. Ask explicitly about the vendor's audio retention policy and ensure it's documented in the BAA.

BAA scope: The BAA must specifically cover the ambient scribing use case — audio capture of encounters. A generic BAA that covers "document processing" may not cover audio capture. Have your compliance officer review the specific BAA language before deployment.

Staff training: All providers using the tool must understand that AI-generated notes require physician review and attestation before signing. The AI generates a draft — the provider is responsible for its accuracy. This is a HIPAA compliance point, not just a quality point.

A practical patient communication script: "I'm using an AI assistant to help with documentation today. It will listen to our conversation to generate my clinical notes. Is that okay with you?"

Most patients say yes. Some say no — in which case, stop the recording and document manually. Have a workflow for both responses.

The First 30 Days: What to Expect

Days 1–3: The tool feels awkward. You're aware of the microphone. You're uncertain whether the AI is capturing everything correctly. This is normal — every provider who now relies on ambient scribing reports an initial adjustment period.

Days 4–14: The workflow becomes natural. You stop thinking about the microphone and start focusing on the patient. Documentation quality from the AI is good on typical encounters; unusual encounters still require more editing.

Days 15–30: You have enough experience to evaluate the tool against your actual workflow. Note quality for your speciality-specific encounter types is calibrated. You know which encounter types require more editing and which produce near-final notes.

The general finding: most ambient scribes consistently save providers 1–2 hours per day on documentation once the tool is calibrated. At typical physician billing rates, even a $500/month AI scribe pays for itself within the first week. At $99/month for Freed or Nabla, the payback is measured in hours.

The most common implementation mistake: Using the tool for one week, deciding it's not working, and stopping. Ambient scribing AI improves with use — it learns your phrasing, your specialty patterns, and your preferred note structure. Give it 30 days before evaluating.

Charles's Recommended Setup by Specialty

Primary care (family medicine, internal medicine, pediatrics): Freed ($99–$149/month) or Nabla (~$119/month paid). Both handle the high-volume, varied encounter types of primary care well. Start with the free tier of either to test before paying.

Mental health (psychiatry, psychology, therapy): Freed or Nabla — both handle the longer, more conversational encounter structure of mental health well. Note: for therapy practices using SimplePractice or Jane App, check whether those platforms' native AI note features (SimplePractice AI Note Taker at $35/month, Jane AI Scribe at $15/month) cover your needs before adding a separate ambient scribing subscription.

Multi-specialty / specialty ambulatory: Suki ($299–$399/month) for its voice command EHR integration that goes beyond documentation, or DeepScribe for specialty-specific model depth. Both require more evaluation of your specific specialty fit.

Dental: Standard dental practice management platforms (Dentrix, Eaglesoft) don't integrate with most ambient scribing tools. For dental clinical notes specifically, the ambient scribing market is less mature — evaluate Jane App's AI Scribe and the native documentation tools in your practice management platform before adding a standalone tool.