A plain-language talk for care teams who want to understand AI before putting it to work. This session builds shared knowledge about draft-making tools, useful tasks, and the checks people need. It is an introductory talk, with no technical background required.
Program abstract
AI is showing up in the tools care teams use each day. Staff may see it in note software, office tools, or chatbots. Knowing what a tool does is the first step toward deciding where it belongs.
In this talk, Cody Saunders, LMSW, explains AI in everyday words. He focuses on generative AI: software that makes new content, such as a text draft, based on learned patterns. Learners see why a clear answer can still contain made-up facts. They compare an office task with a task that supports care. They also learn why a tool approved for one use is not approved for every use.
Made-up examples show how AI might help with repeat writing, note drafts, or missing fields. The aim is less busywork and more room for people. The talk explains how to test those possible benefits by counting review, fixes, and costs too. It does not promise better care, higher billing, or time savings.
Learners leave with a small set of useful terms and one task they can describe for review. They know why real client data needs prior approval and why trained people must check work that affects care. This session gives teams a shared starting point for deeper training and responsible use.
Audience and format
Designed for clinicians, office staff, supervisors, and leaders who are new to AI. Suggested length: 30 minutes, including a short exercise and questions. It can be offered in person or online. No live AI account is needed. This is a proposed talk format, not a claim that the session has already been delivered.
The focus is basic understanding. The separate responsible-use, human-review, and team-rules talks go deeper into work processes. This talk introduces those needs without treating a short session as full training in them.
Learning objectives
After the session, learners should be able to:
- Explain generative AI, a prompt, and an output in plain words. A prompt is the request sent to a tool. An output is the content it returns.
- Spot the difference between clear writing and an answer supported by source facts.
- Name one office use and one care-related use, with a limit for each.
- Describe why the data path and human review need attention before live use.
Proposed session flow
- First 5 minutes: define AI and generative AI. Explain a prompt and an output with a public staff-agenda example.
- Next 7 minutes: compare a public message, a note draft, and a treatment suggestion. Show how the task and data change the review needed.
- Next 8 minutes: complete the made-up draft exercise below. Explain why smooth writing can hide a changed fact.
- Next 5 minutes: describe one task worth testing. Count drafting, review, and fixes when looking for a lighter workload.
- Final 5 minutes: questions and one next step. Point learners to the existing resources for further review.
Sample exercise and debrief
Made-up source: A clinic's public page says its office is open from 9 a.m. to 5 p.m. on weekdays. It does not describe weekend hours.
Made-up AI draft: “The clinic is open every day, including weekends.”
Learners mark what the source supports and what it does not. Then they write a short correction based only on the source.
Answer notes: The draft adds weekend service. A supported rewrite is: “The clinic's posted office hours are 9 a.m. to 5 p.m. on weekdays.” The exercise does not establish whether the clinic offers other services. Learners should not fill a gap with a guess.
The debrief connects this simple error to care records. A tool may also turn missing information into a finding that was never recorded. A person needs the source facts to check meaning. Do not ask learners to share real client notes to prove the point.
Evidence and scope for the presenter
NIST describes confidently stated false content as confabulation, often called an AI hallucination. Its guidance supports explaining why a fluent draft still needs review. It does not establish the error rate of every tool. NIST Generative AI Profile, AI 600-1, July 2024.
For HIPAA-covered uses, cloud providers handling electronic protected health information on an organization's behalf may have business associate duties. Agreements and other applicable safeguards require setting-specific review. Do not tell learners that any named product is approved for client data based on a seller's claim. HHS HIPAA cloud guidance.
These sources support the talk's content; they do not endorse the talk. The sample case, session plan, and learning objectives are proposed teaching materials. No continuing education credit or certification is claimed.
Resources to share after the talk
- What Every Therapist Should Understand About Generative AI: a fuller introduction to draft-making tools.
- AI Hallucinations: Why Behavioral Health Professionals Should Care: how to spot and handle unsupported content.
- AI data-handling worksheet: map data types and review needs without adding client details.
- 25-question purchasing checklist: ask for proof before buying.
Sources checked October 5, 2026. This talk is for learning. It does not replace clinical judgment, legal review, privacy review, or local policy. Use made-up examples and check source updates before delivery.