⚕Competency Health

AI Disclosure

Effective May 1, 2026

Competency Health includes an AI feature called the Competency Coach. This document explains what the Coach does, what it does not do, what data it sees, and how to use it responsibly. Read this before relying on any output the Coach produces.

One-line summary. The Competency Coach is a training-reflection tool. It is not a clinical advisor, a substitute for your preceptor, or a source of medical advice.

1. What the Coach does

The Coach generates short written reflections based on the data you have logged in the Service:

  • A daily summary of skills you logged today and what to focus on next shift
  • A weekly summary of patterns and momentum across your shifts
  • An overall summary spanning your enrollment
  • Focus suggestions identifying competencies you are close to signing off

The Coach is built on a large-language-model API (currently Anthropic's Claude) using a system prompt that scopes its role to training reflection only.

2. What the Coach does NOT do

  • It does not give clinical advice.It will not recommend ventilator settings, interpret a blood gas, suggest a treatment, or weigh in on a bedside decision. If asked, it is instructed to redirect you to your preceptor, charge therapist, or department's protocols.
  • It does not diagnose patients. Do not paste case details, lab values, imaging findings, or patient symptoms.
  • It does not adjudicate sign-off readiness. Sign-offs are decisions made by humans (preceptors, program leadership). The Coach can describe what your data shows; it cannot grant or recommend sign-off.
  • It does not handle emotional, interpersonal, or HR matters. If you raise distress, burnout, harassment, or a conflict with a preceptor, the Coach will respond briefly and direct you to a human resource (your manager, EAP, HR).
  • It does not store, request, or analyze patient information.See “Data the Coach sees” below.

3. Data the Coach sees

When you ask the Coach to generate a summary, the system sends the following to the AI provider:

  • Your first name (no last name, no email)
  • Your current phase, week of enrollment, and recent skill-log entries (competency name, method, confidence level, and any free-text note you typed)
  • Counts of reps, sign-offs, and progress percentages

The Coach is explicitly instructed not to engage with patient-identifying content. If you include patient information in a free-text note, it travels with the request — which is one reason you should not enter PHI into the Service. See Privacy Policy §4.

4. Accuracy and limitations

Large language models can produce confident-sounding output that is wrong, missing context, or stylistically misleading. The Coach is no different. Use it as one input alongside your own judgment, your preceptor's feedback, and your department's protocols.

We tune the Coach for safety (refusing clinical questions, redirecting to humans), but no AI safety system is perfect. If you receive a Coach response that gives clinical advice, contains patient details, or otherwise violates the limits described here, please report it to your manager and to safety@competency.health.

5. Auditing

Every Coach response is logged with a timestamp, your trainee ID, the model version, and the system-prompt version. Your manager and program director can review your Coach history at any time. Behave as if every response will be reviewed — because it may be.

6. Resilience and fallbacks

When the AI provider is unreachable, the Coach falls back to the most recent successful summary and shows a clear “cached” banner. If no cache exists, the Coach generates a deterministic auto-summary built from your activity counts (no AI involved). The fallback is labeled clearly so you know what you're reading.

7. Your responsibilities

  • Treat Coach output as a draft, not as advice. Verify anything that matters with a human.
  • Do not paste patient-identifying information into any free-text field, including Coach prompts.
  • If the Coach goes off-script, redirect — and report the response.

8. Changes

We may update the Coach's system prompt, model, and behavior over time. Significant changes will be announced to administrators and reflected in this disclosure. The current version and effective date are at the top of this page.

9. Contact

Questions about the Coach or AI use: ai@competency.health. Safety concerns: safety@competency.health.