Modern Artificial Intelligence for Medical Practice Teams
What the systems sold as "AI" in a physician practice actually are, and why that changes how you check them. This module separates rules, machine learning and generative AI, explains language models and speech-to-text, opens up retrieval, tools and agents, and maps practice tool categories to the terms, FDA's product line and what physicians report.
-
Rules, Machine Learning and Generative AI in the Practice
35 dk
-
How Language Models and Speech-to-Text Produce Their Output
40 dk
-
Retrieval, Tools and Agents Inside Practice Products
40 dk
-
A Field Guide to Practice AI Tool Categories and Their Vocabulary
45 dk
What AI Gets Wrong in Healthcare Work, and Why It Sounds Right
Why fluent AI output can be wrong in ways that are hard to see. This module covers confabulated rules and citations, stale guidance, omissions in summaries, elaboration on a planted false detail and prompt sensitivity, shows how to read healthcare AI evidence by what each study actually tested, and builds habits that resist automation bias.
-
Confabulated Rules, Citations and Stale Guidance
35 dk
-
Omissions, Planted False Details and Prompt Sensitivity
40 dk
-
Reading Healthcare AI Evidence by What Each Study Tested
40 dk
-
Automation Bias and the Discipline of Doubt
40 dk
Protected Health Information, HIPAA and Where Patient Data Goes
What patient information HIPAA protects, in any form including recordings, and where it goes when someone uses an AI tool. This module covers covered entities and Part 2 records, vendor terms, business associates and their agreements, permitted uses and minimum necessary, real de-identification, and a method for deciding what may go into which system.
-
Covered Entities, Protected Health Information and Part 2 Records
35 dk
-
Vendor Terms, Business Associates and Business Associate Agreements
45 dk
-
Permitted Uses, Minimum Necessary and Real De-identification
45 dk
-
Can I Put This Into This System? A Decision Method
40 dk
Accountability and the Clinical Line: Who Owns What AI Produces
The duties that do not move when an AI tool helps. This module covers the physician accountability that FSMB guidance describes, AMA policy on consent and final review, the line between administrative support and clinical judgment under state examples and Medicare supervision rules, and how staff route every clinical output to a named licensed clinician.
-
The Physician Stays Accountable: FSMB Guidance and What Boards Regulate
35 dk
-
AMA Policy on Consent, Final Review and Disclosure, and What It Binds
35 dk
-
The Administrative Line: Unlicensed Practice, Diagnostic AI and Supervising People
40 dk
-
FDA's Decision Support Line and Routing Every Clinical Output to a Clinician
40 dk
Researching Payer Rules, Regulations and Guidance With AI, and Checking Their Status
How to research a payer rule, regulation or guidance document with AI as an assistant, not an authority. This module builds a research workflow, verifies every citation at its official source, works through current examples of proposed, superseded and vacated rules, and shows how to record the source, the date checked and the status.
-
A Practice Research Workflow From Question to Verified Answer
35 dk
-
Assistant, Not Authority: Verifying Every Citation at the Official Source
40 dk
-
Proposed, Superseded or Vacated: Worked Status Checks on Current Rules
45 dk
-
Courts, Agency Interpretations and Keeping a Dated Research Record
35 dk
Prompt Engineering for Medical Practice Administration
How to write prompts for administrative work in a physician practice, and what prompts cannot do. This module sets out a structured prompt, a data check before any prompt, the limits shown by research on planted false details and the jagged frontier, and draft patterns for schedules, letters, forms, policy summaries, training and translation.
-
The Structured Administrative Prompt: Eight Parts Ending in a Reviewer
35 dk
-
Before the Prompt: Data Class, Minimum Necessary and Fictional Material
35 dk
-
What a Better Prompt Cannot Fix: Planted Details and the Jagged Frontier
40 dk
-
Draft Patterns for Schedules, Letters, Forms, Policy Summaries, Training and Translation
45 dk
Patient Communication With AI Drafts: Messages, Letters, Reminders and Translation
Where AI drafting helps patient communication and where it must stop. This module reads the portal-reply studies by what they measured, routes clinical messages to a licensed clinician, explains California's AB 3030 disclaimer as a state example, and sets limits for reminders, plain-language letters and machine translation.
-
AI-Drafted Portal Replies: What Three Studies Measured and Missed
40 dk
-
Sorting Administrative From Clinical Messages and Routing Them to a Clinician
35 dk
-
Disclosing AI in Patient Messages: California AB 3030 and AMA Policy
35 dk
-
Reminders, Plain-Language Letters and Translation Review
40 dk
Documentation, Ambient Scribes and the Signed Record
How ambient scribes and AI transcription turn a visit into a draft note, and why the draft is not the record until the clinician reviews and signs it. This module covers what the randomised trials found, Medicare's signature rule for AI-transcribed entries, E/M documentation principles, recordings as identifiable data and the staff role.
-
What Ambient Scribes and AI Transcription Do, and What They Do Not
40 dk
-
What the Randomised Ambient Scribe Trials Found
35 dk
-
The Signature on an AI-Transcribed Entry: What Medicare Requires and What It Means
35 dk
-
Reviewing a Draft Note Before Signature: E/M Principles, Note Length and the Staff Role
40 dk
AI in Coding, Billing and the Revenue Cycle
Where AI helps with eligibility, claim scrubbing, denials, appeals and code suggestions, and where it must stop. This module covers the HIPAA code sets and documentation-based coding, the coder's judgment, what three surgical coding studies found, and why unreviewed AI codes can create False Claims Act and overpayment exposure.
-
Where AI Helps in the Revenue Cycle and Where It Must Stop
35 dk
-
Code Sets, Provider Documentation and the Coder's Judgment
40 dk
-
What Three Surgical Coding Studies Found About Language Models
35 dk
-
Knowing Submission, Overpayments and OIG's Billing Risk Areas
45 dk
AI for the Front Office: Scheduling, Phones, Reminders and Records Requests
How to use AI at the front desk without crossing into triage or tripping federal rules. This module covers booking bots and phone agents as administrative support, reminder calls and AI voices under the TCPA, tracking tools and records requests, internal knowledge search, and how to rank front-office uses by risk.
-
Scheduling, Intake and Phone Automation as Administrative Support
40 dk
-
Automated Reminders, Recalls and AI Voices Under the TCPA
40 dk
-
Tracking Tools, Portals and Records Requests at the Digital Front Door
45 dk
-
Front-Office Knowledge Search, Prompt Injection and Ranking Uses by Risk
40 dk
Building Repeatable Practice AI Workflows
How to turn ad hoc AI use into a controlled process a practice can repeat, review and explain. This module builds a six-stage workflow, designs the checks that stop a wrong output becoming a wrong action, strengthens the review step against automation bias, and sets risk-based monitoring and audit records.
-
From Chatbot to a Staged Practice Workflow
40 dk
-
Designing So a Wrong Output Is Not a Wrong Action
40 dk
-
Designing the Review Step Against Automation Bias
40 dk
-
Worked Workflows, Risk-Based Monitoring and Audit Records
45 dk
AI Agents and Autonomous Systems in Practice Operations
What changes when an AI tool can act rather than only answer. This module explains agents and their tools, ranks agent uses from reading to submitting, borrows autonomy vocabulary for control points and stop switches, sets Security Rule expectations for anything touching ePHI, and defends against instructions hidden in incoming documents.
-
What an AI Agent Is, and What Changes When It Can Act
40 dk
-
Agent Risk Levels: Reading, Drafting, Writing, Messaging and Submitting
40 dk
-
Autonomy Vocabulary, Human Control Points and Stop Switches
40 dk
-
Instructions Hidden in Incoming Documents, and Defending Against Excessive Agency
45 dk
Cybersecurity, the HIPAA Security Rule and Vendor Evaluation
How the Security Rule already covers an AI tool that touches electronic patient information: what is required and what is addressable, how the risk analysis is updated when a tool arrives, where the data is actually stored, and how to read a vendor's security claims, authentication and breach reporting terms.
-
The Security Rule as It Stands, and the Rule That Is Only Proposed
40 dk
-
Updating the Risk Analysis When an AI Tool Arrives
40 dk
-
Where the Data Actually Lives: Cloud Hosting, Offshore Storage and a Texas Rule
40 dk
-
Reading Vendor Claims, Choosing Authentication and Knowing the Breach Clock
40 dk
Practice AI Governance and the Compliance Programme
How a practice turns scattered AI use into governance it can explain: a written policy naming approved tools, prohibited data and acceptable use; AI folded into the compliance programme; frameworks placed accurately as voluntary, policy or binding; and the one governance duty that is a regulation, with the records, incidents and review that keep it honest.
-
The Written Practice AI Policy: Approved Tools, Prohibited Data, Acceptable Use
40 dk
-
Fitting AI Into the Practice Compliance Programme
40 dk
-
Placing Every Framework, Policy and Standard by Who It Binds
40 dk
-
The Governance Duty That Binds: Decision Support Tools, Incidents and Review
40 dk
Practical Medical Practice AI Lab
Four hands-on labs on fictional practices and patients: find the planted errors and invented content in an AI-drafted letter, reminder script and payer summary; expose a fabricated citation and rebuild the prompt; review a referral extraction and a failed de-identification; then grade a vendor against business associate requirements and write a tool policy.
-
Lab: Find the Planted Errors in an AI-Drafted Letter, Reminder Script and Payer Summary
45 dk
-
Lab: Expose the Fabricated Citation, Check the Status, Rebuild the Prompt
45 dk
-
Lab: Review a Referral Extraction, Instruction-Like Text and a Failed De-identification
45 dk
-
Lab: Grade a Vendor Against Business Associate Requirements and Write the Tool Policy
45 dk
Capstone: The Medical Practice AI Implementation Plan
Write a sixteen-section implementation plan for one practice task: choose a low-risk, non-sensitive first use with a baseline from your own records, settle the data class, vendor status and risk analysis, set controls, the clinical line and named reviewers, then run a bounded pilot with honest metrics, stop signals and a review cycle.
-
Choosing the First Task and Building a Baseline From Your Own Records
45 dk
-
Data Class, Vendor Status and the Risk Analysis Update in the Plan
45 dk
-
The Clinical Line, Named Reviewers, Training and the Records the Plan Keeps
45 dk
-
The Bounded Pilot, Honest Metrics, Failure Indicators and Improvement
45 dk