Approved workspace rule
Our practice may use Smarter Practice AI for approved dental operations workflows. Patient information and sensitive practice context may only be used in the approved practice workspace.
Policy starter kit
AI policy starter kit for dental teams
Use this starter kit to give the team clear rules before AI use becomes informal, scattered, or unsafe for patient and practice information.
Best for: Owners, office managers, and team leads who need a practical AI policy the team can actually follow.
Asset promise: Practice-owned workflow asset. Use this resource to create one reviewed script, checklist, SOP, packet, scorecard, or decision map the practice can save and reuse.
Build your AI policySee all resources
Sensitive patient, payer, claim, balance, screenshot, and schedule context belongs in approved practice-owned workflows. Use Smarter Practice AI for practice-owned workflows, then let the right team member finish the output before it is used.
Copy/paste starter policy language
Use this as starter language for the owner's next policy and team-training update.
Patient information rule
Team members must use the minimum patient context needed for the task and must not place patient information into personal ChatGPT accounts or unapproved AI tools.
Approved use rule
Approved uses include drafts for scripts, SOPs, claim follow-up, case acceptance follow-up, patient message options, onboarding, huddle notes, and manager summaries.
Prohibited use rule
Prohibited uses include clinical diagnosis, final coding decisions, final insurance or patient-balance decisions, legal advice, HR decisions, and any patient-facing message sent without review.
Practice approval rule
AI output is a draft. The practice must check accuracy, tone, PHI, channel, policy, and next steps before the output is used.
Patient communication rule
Patient-facing drafts must be checked for the correct channel, low-PHI wording, empathy, factual accuracy, and whether a doctor, manager, or billing coordinator should respond.
Clinical and insurance decision limit
AI may help organize questions and draft language, but it may not decide treatment need, urgency, CDT coding, payer coverage, appeal submission, final balance, or contract strategy.
Staff training and acknowledgment rule
Each user must receive AI usage training, acknowledge approved and prohibited uses, and ask the office manager before using AI for a new workflow.
Practice documents to gather before rollout
- Current HIPAA and privacy policies.
- Patient communication rules for text, email, phone, portal, and voicemail.
- Common patient message examples that staff struggle to answer consistently.
- Insurance estimate and balance explanation language currently used by the team.
- SOPs for scheduling, claims, onboarding, daily huddles, and manager review.
- A short list of workflows the owner wants approved first.
Vendor and BAA questions before patient data touches AI
Before patient information or payer-specific context goes into any AI workflow, the owner or privacy lead should confirm whether the vendor creates, receives, maintains, or transmits PHI for the practice, and whether the practice has the right written safeguards in place.
Question to askWill the tool create, receive, maintain, or transmit PHI for the practice?
Why it mattersHHS business associate guidance centers vendor status on whether a service handles PHI for a covered entity or another business associate.
Owner actionTreat patient-specific AI use as vendor review, not just a software feature decision.
Question to askIs there a BAA, and who signs it?
Why it mattersA business associate contract should limit permitted uses, require safeguards, require reporting of unauthorized use, and address subcontractors.
Owner actionConfirm the BAA path before approving patient data, claim notes, EOBs, or patient messages in the tool.
Question to askIs customer data used for model training or unrelated vendor purposes?
Why it mattersThe practice should know whether patient or practice information leaves the approved workflow or is used beyond the service being provided.
Owner actionAsk for the vendor's data-use, retention, deletion, and subcontractor terms in writing.
Question to askWhat user roles, access controls, and audit logs exist?
Why it mattersAI workflows can expose patient, payer, and financial context to more people than intended if access is not managed.
Owner actionDecide who can use each workflow, what gets logged, and who reviews patient-facing or payer-facing output.
Question to askWhat does the AI read or write in the PMS, EHR, or communication system?
Why it mattersDental AI standards and guidance emphasize privacy, security, transparency, safety, and human judgment when AI enters practice workflows.
Owner actionStart with one approved workflow before allowing broad system access or patient-facing automation.
Source note: HHS: Business associate guidance HHS: Business associate contract provisions ADA: Artificial Intelligence in Dentistry
Approved uses
- Drafting patient message options for trained staff review.
- Creating internal call scripts for scheduling, balances, estimates, and follow-up.
- Summarizing claim denial issues and preparing payer call questions.
- Drafting SOPs, onboarding checklists, daily huddle notes, and manager briefs.
- Role-playing patient conversations for training.
- Organizing PPO, expense, schedule, or hygiene questions for owner review.
Prohibited uses
- Putting patient information into personal AI accounts.
- Sending AI-generated patient messages without staff review.
- Asking AI to diagnose, change treatment, or override the doctor.
- Treating AI as final authority on insurance coverage, payer rules, or patient balances.
- Using AI to create legal, tax, contracting, or employment decisions without professional review.
- Copying unnecessary identifiers when a task can be completed with less detail.
PHI and patient-information rules
- Use only the patient context needed for the task.
- Use the approved practice workspace for patient-specific or payer-specific work.
- Keep text, email, and voicemail drafts low-PHI unless the practice's secure-channel rules allow more detail.
- Remove extra identifiers from examples when they do not change the answer.
- Escalate uncertain PHI questions to the office manager or owner before using the tool.
Patient communication and approval rules
AreaTreatment explanation
Review before useDoctor-approved diagnosis, treatment sequence, tone, and patient-specific context.
AI should not decideWhether treatment is clinically necessary or how the doctor should diagnose.
AreaInsurance estimate
Review before useBenefits, frequency limits, deductible, disclaimer, and estimate caveat.
AI should not decideFinal coverage, payer payment, or final patient balance.
AreaBalance explanation
Review before useLedger, EOB, payments, adjustments, pending claims, and collection policy.
AI should not decideFinal amount owed without staff verification.
AreaDifficult patient message
Review before useTone, escalation need, PHI, channel, and whether owner or doctor should respond.
AI should not decideWhether to dismiss, refund, waive, or make a legal-risk statement.
Staff training and manager approval expectations
- New users must receive a short orientation on approved and prohibited AI use.
- The office manager approves new recurring workflows before they become team standard.
- Patient-facing drafts require review by the responsible role before sending.
- Reusable scripts and SOPs should include owner or manager approval date.
- The team should save useful outputs as practice assets instead of one-off private notes.
Team rollout note to send before the meeting
Internal rollout note
Team, before our AI usage meeting, please note that we are moving approved practice workflows into Smarter Practice AI, including scripts, SOPs, claim follow-up, patient message drafts, onboarding, and manager review. Patient information should not be entered into personal ChatGPT accounts. AI output is a draft, not a final answer. Before using anything with a patient, payer, or team member, review it for accuracy, tone, PHI, policy, and the correct next step. If you are unsure whether a use is approved, ask the office manager before using it.
Worked example: patient text about cost, insurance, and pain
Scenario: a front desk team member wants help answering a patient who says the proposed treatment feels expensive, asks whether insurance will cover it, mentions discomfort, and is unsure about scheduling.
MomentWhat should not happen
Policy ruleThe real message, name, balance, appointment date, pain concern, and insurance context should not go into a personal AI account.
Usable outputManager reminds the team to use the approved workspace or ask before using AI.
MomentInside the approved workspace
Policy ruleUse focused context and ask for a patient-safe draft plus internal talking points.
Usable outputDraft response, clinical escalation note, insurance-estimate caveat, and scheduling next-step options.
MomentBefore sending
Policy ruleStaff reviews tone, PHI, pain escalation, estimate caveat, and whether the doctor or coordinator needs to respond.
Usable outputReviewed message and internal note for the patient record or communication log.
MomentAfter approval
Policy ruleSave useful wording as practice guidance, not a one-off private prompt.
Usable outputReusable script for cost-plus-insurance-plus-discomfort messages.
Checklist for reviewing AI-generated output before use
- Does it match practice policy and the actual facts?
- Is the wording appropriate for the channel: phone, text, email, portal, or internal note?
- Does it avoid unnecessary PHI?
- Does it avoid guaranteed insurance, payment, clinical, legal, or financial outcomes?
- Is the responsible staff owner clear?
- Should the output be saved as a reusable script, SOP, or checklist?