Step 1
Decide what AI does
The engine scores intake answers against Aaron's ontology and drafts a first read. Aaron writes the interpretation and the plan. Safety calls, bites and children at risk always go to a person, the same day.
How we use AI
The engine drafts. Aaron decides.
A proposed public statement on how 3D Dog uses AI: where it helps, where it never goes, and who is accountable. Drafted by Citrin Cooperman for Aaron's review.
Step 2
Set it up well
Ask Scout answers only from this site's public pages, with two tools: a link to the assessment and a consult request. Its instructions, tools and limits are written down and published here.
Step 3
Check what it produces
Every AI answer shows its source and carries an "AI-drafted, please check" tag. The first machine read is labeled as a machine read, never as Aaron's interpretation.
Step 4
Govern it
Nothing goes to a client without Aaron's approval. Every submission, AI answer and error is logged. Client answers train nothing outside 3D Dog. AI imagery, if any is ever used, is labeled.
The four steps follow the AI Fluency Framework by Rick Dakan, Joseph Feller and Anthropic (2025, CC BY-NC-SA 4.0), put in our own words.
Where AI shows up
What runs alone, what helps a person, what acts.
| Use | Type | What it does | Status | Human check |
|---|---|---|---|---|
| Assessment scoring | Automates | Maps 41 tapped-and-scored behaviors onto the engine's items and sets a triage tier. | Live | Aaron sees every score |
| First read draft | Assists | Drafts an interpretation for Aaron to edit. Never sent to a client unedited. | Proposed | Aaron approves |
| Safety triage | Automates | Fixed rules flag bites and children at risk for a same-day call. | Live | A person calls |
| Ask Scout | Acts | Answers public questions, sends the assessment link, records consult requests. | Live | Team follows up |
| Learning loop | Assists | Compares Aaron's edits and client accuracy ratings with the engine's read to tune weights. | Proposed | Aaron signs off |
How an assessment flows
From your answers to Aaron's plan.
- IntakeThis site's assessment, server-rendered on Azure FunctionsLive
- EngineThe diagnostic engine scores drivers and safety on Azure FunctionsLive
- DraftAzure OpenAI in Azure AI Foundry writes a first readProposed
- AaronEdits and approves every read and planProposed
- Dynamics 365Each assessment saved as a client, dog and assessment for reviewProposed
- Microsoft FabricOutcomes and accuracy ratings feed the learning loopProposed
Ask Scout
The rules it follows.
- Answers from this site's public pages only, with sources
- Never gives a read, a diagnosis or advice about a specific dog's aggression or bite
- For a bite or a child at risk: call 205-563-8383 today and start the assessment
- Can send you the assessment link or record a consult request for the team
- Logs the topic of a question, never your words