Implementation Guide: Ambient clinical documentation — listen to patient visits, draft soap notes for clinician review
Step-by-step implementation guide for deploying AI to ambient clinical documentation — listen to patient visits, draft soap notes for clinician review for Healthcare clients.
Hardware Procurement
Philips SpeechMike Ambient
$380–$420 per unit (MSP cost) / $499–$549 suggested resale
Professional-grade wearable ambient capture microphone with four beamforming microphones, patented AI speaker-separation technology, and selectable audio modes for doctor-patient conversations. Provides significantly better audio quality than smartphone mics in noisy clinical environments, directly improving transcription accuracy.
Philips SpeechMike Premium Touch (desktop fallback)
Philips SpeechMike Premium Touch
$300–$400 per unit (MSP cost) / $399–$449 suggested resale
USB desktop dictation microphone for providers who prefer a desk-based workflow or for telehealth encounters where a wearable mic is unnecessary. Compatible with Dragon Medical One and all browser-based ambient scribes. Keep 1–2 as spares or for telehealth-heavy providers.
Apple iPad 10th Generation Wi-Fi 64GB
$449 per unit (MSP cost) / $549–$599 suggested resale
Dedicated ambient capture tablet for exam rooms where clinicians prefer not to use personal devices. Required for DAX Copilot (iOS only). Also serves as a secondary capture device for Freed AI, Twofold, or Sunoh mobile apps. Pre-configured with MDM profile.
Jabra Speak 750 USB/Bluetooth Speakerphone
$200–$250 per unit (MSP cost) / $299–$349 suggested resale
Conference-style room microphone for larger exam rooms, group therapy rooms, or dental operatories where wearable mics are impractical. 360-degree pickup pattern captures multi-speaker conversations. USB and Bluetooth connectivity.
Ubiquiti UniFi U6 Enterprise Access Point
$329 per unit (MSP cost) / $449–$499 suggested resale
Wi-Fi 6E access points to ensure reliable, low-latency wireless coverage in all exam rooms and operatories. Required for real-time audio streaming to cloud AI services. Deploy one per floor or wing to ensure <50ms latency to cloud endpoints. Supports WPA3-Enterprise and VLAN tagging for clinical device segmentation.
Ubiquiti UniFi Dream Machine Pro
$379 per unit (MSP cost) / $499–$549 suggested resale
All-in-one gateway, switch, and network controller for VLAN segmentation (clinical devices, guest, corporate), firewall rules, and centralized UniFi management. If client already has a managed firewall (SonicWall, Meraki, Fortinet), skip this and configure VLANs on existing infrastructure.
OtterBox Defender iPad Case with Hand Strap
$70 per unit (MSP cost) / $99 suggested resale
Rugged protective case for iPads used in clinical environments. Prevents drops and allows easy single-hand operation during patient encounters. Includes screen protector for infection control wipe-downs.
Kensington USB-C Docking Station
$180 per unit (MSP cost) / $249 suggested resale
Thunderbolt/USB-C dock for clinician workstations to connect Philips SpeechMike USB microphones, external monitors, and wired Ethernet. Ensures reliable USB audio connection for desktop-based ambient capture workflows.
Software Procurement
Freed AI Premier
$119/month per provider (monthly) or $104/month per provider (annual billing). Pass-through to client with $20–$30/month MSP management surcharge.
Primary ambient clinical documentation platform for medical practices. Listens to patient encounters via mobile app or desktop, generates SOAP notes using AI, and pushes completed notes to the EHR via Chrome browser extension. Includes EHR push integration, visit summaries, patient context, and instant patient instructions. No API development required.
Twofold Health
$69/month (monthly) or $49/month billed annually ($588/year)
Budget-tier ambient clinical documentation platform supporting SOAP, DAP, and BIRP note formats. Includes unlimited notes, custom templates, treatment plans, and mobile/desktop apps. Fully HIPAA and HITECH compliant with BAA provided at signup. Recommended for behavioral health, therapy, and budget-constrained small practices. Budget alternative to Freed.
Bola AI
Estimated $200–$400/month per dental location. Contact vendor for MSP partner pricing.
Primary ambient clinical documentation platform for dental practices. Purpose-built for dental workflows including perio charting, restorative procedures, and dental terminology. Direct PMS integration with Dentrix, Eaglesoft, and Open Dental. Trusted by over 10,000 dentists and hygienists. Voice Perio + Clinical Notes modules.
DentScribe by DeepCura
$129/month per provider. Alternative to Bola AI for dental.
Dental ambient scribe that integrates with Dentrix, Eaglesoft, and Open Dental. Notes are automatically published to the PMS when the appointment ends. Good alternative if Bola AI pricing is prohibitive or if practice uses a PMS not supported by Bola.
Google Chrome Browser
$0
Required browser for Freed AI's EHR push Chrome extension. Must be latest stable version on all clinician workstations. Extension enables one-click note push to web-based EHRs (athenahealth, eClinicalWorks, DrChrono, Practice Fusion, AdvancedMD).
Mosyle Business (Apple MDM)
$4–$8/month per device depending on tier
Mobile device management for practice-owned iPads and iPhones used for ambient capture. Enables remote configuration, app deployment, device lock, remote wipe, and compliance enforcement. Essential for HIPAA device management requirements.
Huntress Managed EDR
$4–$6/month per endpoint (MSP cost); bill client $8–$12/endpoint
Managed endpoint detection and response for all clinical workstations. Required for HIPAA Security Rule compliance — provides audit logging, threat detection, and incident response capabilities. Monitors workstations where PHI-containing notes are accessed.
Vanta or Secureframe (HIPAA Compliance Platform)
$5,000–$10,000/year for the practice (or use MSP's existing compliance platform and bill client a portion)
Automated HIPAA compliance monitoring, risk assessment documentation, BAA tracking, policy management, and evidence collection. Generates the required HIPAA Security Risk Assessment that must now include AI tools per proposed 2025 HHS regulation. Can be shared across multiple MSP clients to reduce per-client cost.
Prerequisites
- Active EHR subscription: Medical practices must have a functioning cloud-based or web-accessible EHR system (e.g., athenahealth, eClinicalWorks, DrChrono, AdvancedMD, Practice Fusion, Epic). Dental practices must have an active PMS (Dentrix, Eaglesoft, Open Dental, Curve Dental, or CareStack).
- Internet connectivity: Minimum 25 Mbps symmetrical broadband for a 5-provider practice. Each concurrent ambient capture stream requires approximately 256 kbps upload, but total practice bandwidth must support simultaneous EHR, imaging, and ambient AI traffic.
- Wi-Fi coverage: Reliable wireless signal (minimum -65 dBm) in every exam room, operatory, and clinical area where patient encounters occur. If existing coverage is inadequate, include Wi-Fi 6E access point deployment in project scope.
- Clinician devices: Each provider needs either (a) a personal or practice-owned smartphone (iOS 15+ or Android 10+) or (b) a Philips SpeechMike Ambient wearable microphone connected to their workstation. iPads are optional but recommended for shared-device workflows.
- Chrome browser: Latest stable version of Google Chrome installed on all clinician workstations for browser-extension-based EHR integration (Freed AI, Twofold).
- HIPAA compliance foundation: Practice must have an existing or concurrent HIPAA Security Risk Assessment, documented policies and procedures, and workforce training. The MSP must update the risk assessment to include ambient AI tools per the 2025 HHS proposed regulation.
- Legal review of state recording consent laws: Before deployment, the practice's legal counsel must confirm that the recording consent approach (written intake form + verbal confirmation + room signage) meets the requirements of the state where the practice operates. This is critical in all-party consent states: CA, CT, DE, FL, IL, MD, MA, MI, MT, NV, NH, OR, PA, VT, WA.
- Business Associate Agreement: The practice (covered entity) must execute a BAA with the AI scribe vendor (business associate) before any PHI is processed. Most SMB vendors (Freed, Twofold, Bola) provide click-through BAAs during signup; document execution date and retain a copy.
- Administrative access: MSP technician requires admin credentials for the practice's EHR/PMS system, network infrastructure (firewall, switches, APs), and any MDM platform. Obtain these during the pre-engagement scoping call.
- Clinician availability: Schedule 2-hour blocks for each provider for initial training and template customization. Plan for reduced patient volume on go-live day (recommend 50% schedule for first half-day).
Installation Steps
Step 1: Network Assessment and Wi-Fi Optimization
Perform a wireless site survey of all clinical areas to ensure adequate Wi-Fi coverage for real-time audio streaming. Ambient capture requires consistent, low-latency connectivity in every room where patient encounters occur. Use a Wi-Fi analyzer tool to measure signal strength, channel utilization, and interference in each exam room and operatory. Target minimum -65 dBm signal strength and <50ms latency to the vendor's cloud endpoints.
iperf3 -c iperf.he.net -p 5201 -t 30ping -c 100 api.freed.ai/System/Library/PrivateFrameworks/Apple80211.framework/Versions/Current/Resources/airport -I | grep agrCtlRSSInetsh wlan show interfaces | Select-String 'Signal'If any exam room shows signal weaker than -70 dBm, add a Ubiquiti U6-Enterprise AP in that zone. For practices in older buildings with plaster/lath walls, consider one AP per 2–3 rooms. Document all readings in a site survey report for the client file.
Step 2: VLAN Segmentation for Clinical Devices
Create a dedicated VLAN for clinical ambient capture devices (iPads, smartphones, SpeechMike-connected workstations) to isolate PHI-carrying traffic from guest and general office networks. This is a HIPAA Security Rule requirement for network segmentation. Configure firewall rules to allow only HTTPS (443) and DNS (53) outbound from the clinical VLAN, and block inter-VLAN routing to the guest network.
If the practice already has a managed firewall (SonicWall, Meraki, Fortinet, etc.), create the VLANs on that platform instead. The key requirement is that clinical devices carrying audio PHI are on an isolated network segment. Assign the clinical SSID (e.g., 'ClinicSecure') to VLAN 10 with WPA3-Enterprise or WPA2-Enterprise authentication.
Step 3: Configure and Deploy Microphone Hardware
Unbox, configure, and deploy Philips SpeechMike Ambient (PSM5000) wearable microphones for each provider. Connect desktop SpeechMike Premium Touch (SMP3800) units to workstations via USB docking stations for providers who prefer desk-based capture. Verify audio input levels and test speaker separation in a simulated two-person conversation.
SpeechControl_Setup.exe /S /v/qnGet-PnpDevice -Class AudioEndpoint | Where-Object {$_.FriendlyName -like '*Philips*'}The PSM5000 wearable mic should be worn on the clinician's lapel or lanyard at chest height. Test speaker separation by having two people speak simultaneously from 3–6 feet apart — the AI should correctly attribute speech to different speakers. If using Jabra Speak 750 for room-based capture, place it centrally on the exam table or counter, equidistant from provider and patient seating positions. Label each microphone with the provider's name and room number.
Step 4: MDM Enrollment for Practice-Owned Mobile Devices
Enroll all practice-owned iPads and iPhones into the Mosyle MDM platform. Configure device profiles to enforce encryption, passcode requirements, auto-lock timeout (2 minutes maximum for HIPAA), and approved app deployment. Push the ambient AI app (Freed, Twofold, or Bola) to all enrolled devices. Restrict non-clinical app installation to prevent unauthorized software.
If clinicians use personal (BYOD) devices, skip MDM enrollment but document a BYOD policy that requires device encryption, passcode, and current OS version. The BYOD policy should be signed by each clinician. For BYOD scenarios, the ambient AI app itself provides data isolation — audio is streamed directly to the vendor cloud and not stored locally on the device.
Step 5: Create Vendor Accounts and Execute BAAs
Create the practice's account on the chosen ambient AI platform (Freed AI for medical, Bola AI or DentScribe for dental). During signup, execute the Business Associate Agreement — this is a legal prerequisite before any PHI is processed. Create individual provider sub-accounts under the practice master account. Record BAA execution date, signatory, and retain a PDF copy in the compliance file. ``` # Freed AI Signup (Medical) # 1. Navigate to https://app.freed.ai/signup # 2. Select 'Premier' pla...
Step 6: Install and Configure Chrome Extension for EHR Integration (Medical)
Install the Freed AI Chrome extension on each clinician's workstation. This extension enables one-click push of AI-generated SOAP notes directly into the practice's web-based EHR (athenahealth, eClinicalWorks, DrChrono, Practice Fusion, AdvancedMD). Configure the extension to recognize the EHR's note entry fields and map them to the AI-generated note sections (Subjective, Objective, Assessment, Plan).
Install Freed AI Chrome Extension
Deploy via Group Policy to All Workstations (Windows Registry / RMM / GPO)
reg add "HKLM\SOFTWARE\Policies\Google\Chrome\ExtensionInstallForcelist" /v 1 /t REG_SZ /d "<FREED_EXTENSION_ID>;https://clients2.google.com/service/update2/crx" /fThe Chrome extension approach does NOT use EHR APIs — it works by interacting with the EHR's web interface through the browser. This means it works with virtually any web-based EHR without vendor API credentials. However, it requires Chrome to be the clinician's primary browser. If the EHR uses a thick client (installed application, not browser-based), the Chrome extension won't work — in that case, clinicians will need to use the copy-paste workflow from the Freed web/mobile app into the EHR desktop application.
Step 7: Configure Dental PMS Integration (Dental Practices)
For dental practices using Bola AI or DentScribe, configure the direct integration with the practice management system (Dentrix, Eaglesoft, or Open Dental). This integration enables AI-generated clinical notes and perio charting data to write back directly into the patient chart without copy-paste. The integration method varies by PMS: Dentrix and Eaglesoft typically use a local integration agent, while Open Dental uses its REST API.
Bola AI + Dentrix Integration
Bola AI + Open Dental Integration
DentScribe + Eaglesoft Integration
Alternative: Sikka ONE API (middleware for multi-PMS environments)
Contact Sikka Software: https://www.sikkasoft.com — Sikka provides a single REST API that connects to 30+ dental PMS systems.
Dental PMS integrations are more complex than medical EHR browser extensions because dental PMS systems (especially Dentrix and Eaglesoft) are often locally installed server-client applications, not cloud/web-based. The integration agent must run on the same machine or network as the PMS server. Coordinate with the dental practice's PMS support vendor to avoid conflicts. Schedule integration testing during non-patient hours. The Bola AI onboarding team typically handles most of the configuration — the MSP's role is to provide network access, server credentials, and be present for troubleshooting.
Step 8: Customize Note Templates and Clinical Workflows
Configure AI note templates to match each clinician's documentation style and specialty. Most ambient AI platforms come with default SOAP templates, but customization is critical for clinician adoption. Work with each provider to define their preferred note structure, included/excluded sections, common diagnoses terminology, and any practice-specific documentation requirements. For dental, configure templates for different procedure types (perio, restorative, hygiene, oral surgery).
Freed AI Template Customization
Bola AI Dental Template Customization
Template customization is the #1 factor in clinician satisfaction with ambient AI scribes. Allocate at least 30 minutes per provider for template review and customization. Encourage providers to use the default templates for the first week, then refine based on actual generated notes. Both Freed and Twofold improve note quality over time as the AI learns the clinician's style and preferences. Save template configurations in your MSP documentation so they can be restored if the provider's account is ever recreated.
Step 9: Implement Patient Consent Workflow
Design and deploy a patient consent process for audio recording of clinical encounters. This is legally required in all-party consent states and strongly recommended in all states as a HIPAA best practice. The workflow includes three components: (1) written consent on the patient intake form, (2) visible signage in all exam rooms and operatories, and (3) verbal confirmation at the start of each recorded encounter, preferably captured by the recording itself.
CRITICAL LEGAL NOTE: In all-party consent states (CA, CT, DE, FL, IL, MD, MA, MI, MT, NV, NH, OR, PA, VT, WA), recording without consent may be a criminal offense. The practice's legal counsel MUST review the consent language before deployment. For dental practices, add consent to the standard dental intake packet. For pediatric patients, the parent/guardian must sign consent. Keep signed consent forms on file as part of the patient record. If a patient opts out, the clinician must not activate the recording for that encounter — ensure the workflow supports this opt-out gracefully.
Step 10: Clinician Training and Supervised Go-Live
Conduct hands-on training sessions with each clinician and relevant support staff (medical assistants, dental assistants, hygienists, front desk). Cover the complete workflow: activating recording, speaking naturally during the encounter, reviewing the AI-generated draft note, making corrections, and pushing the final note to the EHR/PMS. Shadow the first 5–10 real patient encounters per clinician to provide real-time support and troubleshoot issues.
Training Agenda (30–45 minutes per provider)
1. Overview (5 min)
- What ambient AI scribing does and doesn't do
- Clinician remains responsible for note accuracy (AI generates DRAFT only)
- Note review and approval is mandatory before signing/committing
2. Device Setup (5 min)
- Open Freed/Twofold/Bola app on phone or start desktop app
- Verify microphone is active (PSM5000 wearable or SMP3800 desktop)
- Confirm Wi-Fi connection on clinical VLAN
3. Recording Workflow (10 min)
- Tap 'Start Visit' or 'Record' at the beginning of the encounter
- Speak naturally — no special voice commands or dictation style needed
- The AI distinguishes clinician from patient speech automatically
- Tap 'End Visit' when the encounter concludes
- Wait 30–90 seconds for AI note generation
4. Note Review & Editing (10 min)
- Open the generated draft note in the app
- Review each section (S, O, A, P) for accuracy
- Edit any errors, add missing details, remove hallucinated content
- Approve the note when satisfied
5. EHR Push (5 min)
- Click the Chrome extension (Freed) or confirm PMS write-back (Bola)
- Verify the note appears correctly in the patient chart
- Sign/finalize the note per normal EHR workflow
6. Opt-Out Handling (5 min)
- If a patient declines recording: simply don't start the recording
- Document the visit manually as before
- No patient data is ever captured without active recording initiation
Supervised Go-Live Checklist
Schedule go-live during a lighter patient day (e.g., half-day or afternoon-only session). The MSP technician should be physically present at the practice for at least the first half-day of go-live. Common first-day issues include: Wi-Fi drops in specific exam rooms (move AP or add repeater), microphone placement too far from speakers (adjust wearable position), AI misattributing speech between provider and patient (retrain speaker profiles), and note format not matching clinician expectations (adjust templates). Most clinicians are comfortable self-managing by day 3–5.
Step 11: HIPAA Security Risk Assessment Update
Update the practice's HIPAA Security Risk Assessment to include the new ambient AI documentation system. Per the 2025 HHS proposed regulation, entities using AI tools must include those tools as part of their risk analysis and risk management compliance activities. Document the AI vendor, data flows, security controls, BAA status, and residual risks. This is a billable compliance service that the MSP should perform annually. ``` # Risk Assessment Documentation Template (store in Vanta/Securefra...
Step 12: Post-Deployment Monitoring and 30-Day Optimization
Monitor the deployment for the first 30 days to ensure stable operation, clinician adoption, note quality, and system reliability. Conduct weekly check-ins with the practice during this period. Track key metrics: notes generated per provider per day, note approval rate (approved without edits vs. edited before approval), clinician time savings, and any technical issues. Adjust templates, microphone placement, and workflows based on real-world usage data. ``` # Weekly Check-in Agenda (15 min pho...
Custom AI Components
Patient Consent Tracking Workflow
Type: workflow An automated workflow that tracks patient recording consent status in the EHR scheduling system, alerts front desk staff when a patient has not yet signed the AI recording consent form, and flags opted-out patients to prevent accidental recording. This is critical for legal compliance in all-party consent states. Implementation: ``` ## Patient Consent Tracking Workflow ### Overview This workflow uses the practice's existing EHR/PMS and a simple tracking mechanism to en...
SOAP Note Quality Validation Prompt
Type: prompt
A quality assurance prompt template that clinicians can use to validate AI-generated SOAP notes before signing. The prompt checks for common AI hallucinations, missing required elements, and documentation compliance issues. Can be used manually or integrated into a GPT-based QA workflow.
Purpose
This prompt is designed to be used as a secondary AI check on ambient scribe-generated SOAP notes. A clinician or QA staff member can paste the AI-generated note into ChatGPT (with appropriate HIPAA considerations — use de-identified data only, or use Azure OpenAI with BAA) to check for common documentation issues.
Prompt Template
# paste into ChatGPT or Azure OpenAI with a generated note to receive a
# structured quality assessment
You are a clinical documentation quality assurance specialist. Review the following SOAP note draft generated by an AI ambient scribe for a [SPECIALTY] encounter. Evaluate it against these criteria and provide a structured assessment:
## EVALUATION CRITERIA
1. **COMPLETENESS CHECK**
- Does the Subjective section include: Chief complaint, HPI (onset, location, duration, character, aggravating/alleviating factors, timing, severity), relevant ROS, relevant PMH/PSH/medications/allergies/social history?
- Does the Objective section include: Vitals (if applicable), physical exam findings relevant to the chief complaint, any diagnostic results discussed?
- Does the Assessment section include: A clear diagnosis or differential, supported by the subjective and objective findings?
- Does the Plan section include: Treatment plan, medications (with dose/route/frequency if discussed), follow-up timeline, patient education, referrals?
2. **HALLUCINATION CHECK**
- Flag any clinical details that seem unusually specific but were likely not explicitly stated (e.g., exact lab values, specific medication doses, detailed exam findings that would require physical examination)
- Flag any diagnoses that don't logically follow from the documented symptoms and exam
- Flag any medications listed in the plan that weren't mentioned in the conversation context
3. **ATTRIBUTION CHECK**
- Is patient-reported information correctly attributed to the patient (Subjective) vs. clinician findings (Objective)?
- Are any clinician statements incorrectly recorded as patient statements, or vice versa?
4. **COMPLIANCE CHECK**
- Does the note support the likely billing level (E&M code)?
- Are required elements for the likely code level documented?
- Any sensitive information (substance use, mental health, STI) that requires special handling?
## SOAP NOTE TO REVIEW:
[PASTE AI-GENERATED NOTE HERE]
## OUTPUT FORMAT:
Provide a structured assessment with:
- Overall Quality Score (1-5)
- Completeness Issues (list)
- Potential Hallucinations (list with reasoning)
- Attribution Errors (list)
- Compliance Concerns (list)
- Recommended Edits (list)Usage Notes
HIPAA WARNING: Do NOT paste notes containing real PHI into consumer ChatGPT. Use only with: de-identified notes (remove patient name, DOB, MRN), Azure OpenAI Service with an executed BAA, or an on-premises LLM deployment.
- This prompt is most useful during the first 2 weeks of deployment when calibrating AI note quality.
- After the initial period, clinicians develop their own quick-review habits and this formal QA becomes optional.
- For practices requiring formal QA (teaching hospitals, compliance-sensitive environments), consider building this into a dedicated Azure OpenAI workflow.
Ambient Scribe Performance Dashboard
Type: integration A lightweight monitoring dashboard that tracks ambient AI scribe adoption metrics, note quality indicators, and cost-per-note economics across the practice. Built using the MSP's existing PSA/reporting tools or a simple Google Data Studio dashboard, pulling data from vendor usage reports and clinician feedback.
Data Sources
Dashboard Metrics
Sheet 1: Weekly Adoption Metrics
| Metric | Week 1 | Week 2 | Week 3 | Week 4 | Target |
|---------------------------------|--------|--------|--------|--------|----------|
| Total encounters recorded | | | | | 80%+ eligible |
| Encounters per provider/day | | | | | Match patient volume |
| Active providers / Licensed | | | | | 100% |
| Patient opt-out rate | | | | | <5% |
| Notes approved without edits | | | | | >60% |
| Notes requiring major edits | | | | | <10% |
| Avg time savings per encounter | | | | | >3 min |
| EHR push success rate | | | | | >95% |
| Technical issues reported | | | | | <3/week |Sheet 2: ROI Calculator
| Item | Value |
|-----------------------------------------|------------|
| Number of providers | [N] |
| Monthly AI scribe cost | [N × $119] |
| Monthly MSP managed services | [$350] |
| Total monthly cost | [SUM] |
| Avg minutes saved per encounter | [X] |
| Avg encounters per provider per day | [Y] |
| Total minutes saved per month | [CALC] |
| Provider hourly rate | [$Z] |
| Monthly time-value recovered | [CALC] |
| Additional patients seen (est.) | [CALC] |
| Additional revenue per patient (est.) | [$150-$300]|
| Net monthly ROI | [CALC] |Automation
- Set up a monthly automated email (via Google Sheets scheduled email or PSA tool) to the practice manager with the dashboard summary.
- Include month-over-month trends and highlight any adoption or quality issues.
- This dashboard is a key deliverable for justifying the ongoing MSP managed services fee.
Implementation Steps
Encounter Recording State Machine
Type: workflow A defined state machine that governs the lifecycle of each ambient-recorded encounter, from consent verification through note approval and EHR commit. Ensures no encounter falls through the cracks and provides an auditable trail for compliance purposes.
States
[SCHEDULED] → [CHECKED_IN] → [CONSENT_VERIFIED] → [RECORDING_ACTIVE] → [RECORDING_COMPLETE] → [NOTE_GENERATED] → [CLINICIAN_REVIEW] → [NOTE_APPROVED] → [EHR_COMMITTED] → [COMPLETE]
↓ ↓
[CONSENT_DECLINED] [NOTE_REJECTED]
↓ ↓
[MANUAL_DOCUMENTATION] [MANUAL_DOCUMENTATION]State Definitions
AI_Recording_Consent fieldAI_Recording_Consent = Consented
- *Action*: No flag needed; clinician may proceed with recordingAI_Recording_Consent = Declined or Pending
- *Action*: Flag in schedule: 'No AI Recording'. Clinician documents manually.Implementation Notes
This state machine is conceptual — it is NOT implemented as software code but as a documented workflow that is trained to all staff. The 'states' map to real actions taken by front desk, clinician, and AI system.
- Print this workflow as a laminated reference card for each exam room during the first month.
- Use the states to troubleshoot where notes are getting 'stuck' (e.g., generated but not reviewed = clinician adoption issue; reviewed but not committed = EHR push technical issue).
- Track notes in each state during the 30-day optimization period to identify bottlenecks.
Testing & Validation
Client Handoff
Client Handoff Meeting Agenda (60–90 minutes with Practice Owner/Manager and Lead Clinician)
1. System Overview & Architecture (10 min)
- Review the deployed solution: which AI platform, which devices, how data flows from microphone → cloud → EHR
- Show the network diagram: clinical VLAN, Wi-Fi APs, internet path
- Explain what is cloud-hosted (AI processing) vs. local (devices, EHR)
2. Account Credentials & Access (10 min)
- Hand over master admin credentials for the AI scribe platform (Freed/Bola/DentScribe) - Document all accounts in a secure password manager (recommend practice adopt Keeper or 1Password with BAA) - Review MDM admin access (Mosyle) and who can manage devices - Provide list of all licensed providers and their account status...
3. Daily Workflow Review (15 min)
- Walk through the complete workflow: patient check-in → consent verification → encounter recording → note review → EHR push → note signing
- Review the state machine diagram and laminated reference cards in exam rooms
- Confirm front desk staff can demonstrate the consent workflow independently
- Confirm each clinician can demonstrate the complete recording-to-EHR workflow independently
4. Compliance Documentation (15 min)
5. Performance Metrics Review (10 min)
- Present the 30-day performance dashboard showing adoption rates, note quality scores, and time savings
- Calculate and present ROI: cost of the solution vs. estimated value of clinician time recovered and additional patients seen
- Discuss any providers with low adoption and remediation plan
6. Ongoing Support & Escalation (10 min)
- Review the monthly managed services agreement scope: what's included (template updates, Chrome extension monitoring, quarterly compliance review, vendor liaison) vs. what's billable separately (new provider onboarding, network changes, major EHR upgrades)
- Provide MSP help desk contact info and expected response times
- Explain escalation path: L1 (MSP help desk) → L2 (MSP senior tech) → L3 (vendor support)
- Provide direct vendor support contact info as backup
7. Documentation Package Left Behind
8. Success Criteria Sign-Off
Maintenance
Ongoing Managed Services Responsibilities
Monthly Tasks (included in $200–$500/month managed services fee)
- AI Platform Health Check: Log into vendor admin portal, verify all provider accounts are active, check for any error alerts or degraded performance notices. Confirm license count matches active providers.
- Chrome Extension Update Verification: Ensure the Freed AI Chrome extension is up-to-date on all clinician workstations. Chrome auto-updates extensions, but verify via RMM that no workstations have disabled or removed the extension.
- Network Monitoring: Review clinical VLAN connectivity, Wi-Fi AP uptime, and any alerts from the UDM-Pro or managed firewall. Ensure no rogue devices have joined the clinical VLAN.
- Device Compliance Check: Review MDM dashboard (Mosyle) to confirm all practice-owned devices remain compliant (encryption, passcode, OS updates, approved apps only).
- Performance Dashboard Update: Populate the monthly performance dashboard with adoption metrics from the vendor portal and any clinician feedback. Send monthly summary report to practice manager.
Quarterly Tasks
- Template Review: Schedule a 15-minute call with 1–2 clinicians to review note template satisfaction. Make adjustments as needed based on evolving documentation requirements, payer changes, or clinician preferences.
- Compliance Spot-Check: Verify patient consent forms are still being collected (spot-check recent patient charts for consent documentation). Confirm BAAs are current and no vendor changes have occurred.
- Vendor Relationship Check: Review vendor communications for product updates, pricing changes, new features, or deprecation notices. Communicate relevant changes to the practice.
- Security Review: Run vulnerability scan on clinical workstations. Review Huntress alerts for the practice. Verify no security incidents involving the AI platform or clinical devices.
Annual Tasks
- HIPAA Security Risk Assessment Update: Conduct full annual update to the risk assessment, including the AI tool addendum. Bill separately at $500–$1,000.
- Patient Consent Re-authorization: Reset all patient consent statuses to 'Pending' and notify front desk to re-collect consent during January patient visits.
- License True-Up: Review vendor licenses vs. active providers. Add or remove seats as needed. Renegotiate annual pricing if applicable.
- Hardware Assessment: Inspect microphones, iPads, and APs for wear and damage. Replace batteries (if applicable) and refresh devices approaching end-of-life.
- Vendor Contract Renewal: Review AI platform subscription terms, pricing, and BAA before annual renewal. Negotiate pricing improvements if volume has increased.
Escalation Paths
- L1 — MSP Help Desk (response within 2 hours during business hours): Basic troubleshooting — app not recording, Chrome extension not pushing, Wi-Fi connectivity, device issues.
- L2 — MSP Senior Technician (response within 4 hours): EHR integration failures, network/VLAN issues, MDM problems, template reconfiguration, new provider onboarding.
- L3 — Vendor Support (response per vendor SLA): AI note quality degradation, platform outages, API/integration failures, billing disputes. MSP acts as liaison between practice and vendor.
- Emergency — Data Breach / HIPAA Incident: Follow practice's incident response plan. MSP assists with technical investigation, containment, and evidence preservation. Practice's legal counsel manages breach notification per HIPAA Breach Notification Rule (60-day window for breaches affecting 500+ individuals).
SLA Considerations
- AI scribe platform availability: Dependent on vendor SLA (typically 99.5–99.9% uptime). MSP cannot guarantee vendor uptime but commits to rapid escalation.
- MSP response time: 2-hour response during business hours (M-F 8am-6pm), next business day for non-urgent after-hours requests.
- Note generation: Most platforms generate notes within 30–90 seconds. If consistently exceeding 3 minutes, escalate to vendor.
- EHR push reliability: Target 95%+ success rate. Investigate if failure rate exceeds 5% in any given week.
Model/AI Update Considerations
- AI scribe vendors regularly update their language models. These updates are vendor-managed and transparent to the practice.
- After major vendor model updates, monitor note quality closely for 1–2 weeks (model updates can temporarily change note style or introduce new behaviors).
- Subscribe to vendor release notes/changelogs and brief the practice on significant changes.
- No MSP-side model retraining is required — all AI training/tuning is vendor-managed in the SaaS model.
Alternatives
Nuance DAX Copilot (Enterprise Tier)
Microsoft/Nuance Dragon Ambient eXperience (DAX) Copilot is the market-leading enterprise ambient clinical documentation platform. It requires a Dragon Medical One subscription as a prerequisite and offers the deepest EHR integration with Epic, including AI-generated order suggestions within the Epic workflow. Priced at $600/month per user with a $650 one-time setup fee for the first user. Reseller pricing via TryDAX starts at $369/month with $700 implementation fee per user. iOS-only mobile app...
Twofold Health (Budget Tier)
Twofold Health is the lowest-cost ambient clinical documentation platform at $49/month billed annually per provider with unlimited notes. Supports SOAP, DAP, and BIRP note formats. Includes mobile and desktop apps with HIPAA/HITECH compliance and a BAA provided at signup. Best suited for behavioral health, therapy practices, and budget-constrained small practices. Tradeoffs: At $49/month vs. Freed's $119/month, Twofold saves $840/year per provider. However, Twofold's EHR integration is less...
Suki AI (Mid-Market Voice Assistant)
Suki AI is a voice-first clinical assistant priced at $299–$399/month per provider. It goes beyond ambient documentation to include voice commands for EHR navigation ('Suki, show me the latest labs'), prescription entry, and order placement. Integrates with 100+ EHR systems via API. More powerful than pure scribes but with a steeper learning curve. Tradeoffs: Higher cost ($299–$399/mo vs. $119/mo for Freed) and longer implementation timeline (2–6 weeks vs. 1–3 days). However, Suki provides ...
DeepScribe (Hybrid AI + Human QA)
DeepScribe uses a hybrid approach combining AI-generated notes with human quality assurance review. Priced at $400/month (non-EHR integrated) or $500/month (EHR integrated) with up to 20% discounts for annual commitments. Integrates with athenahealth, eClinicalWorks, Epic, and AdvancedMD.
VideaHealth Voice Notes (Dental AI Platform)
VideaHealth is the first AI-powered ambient scribe created specifically for dentistry, launched October 2025. It achieves 95% complete clinical notes on the first pass and is part of the broader VideaAI platform that includes FDA-cleared diagnostic imaging AI. Trusted by 8 of the 10 largest DSOs.
Open-Source / Self-Hosted Approach (Whisper + LLM)
Build a custom ambient documentation pipeline using OpenAI Whisper (open-source speech-to-text) for transcription and an open-source LLM (Llama 3, Mistral) for SOAP note generation. Host on-premises or in a private cloud to maintain full data sovereignty. Requires GPU hardware (NVIDIA A100 or equivalent), custom development, and ongoing maintenance.
Want early access to the full toolkit?