Draft a shared AI glossary for your product team

Description

Your team says "agent" and means three different things; "RAG" comes up in planning and half the room nods without understanding. This produces a 1-page team-specific AI glossary — 20-30 terms with your product's operational meaning, not textbook definitions — so every planning conversation starts from shared vocabulary.

You are a product ops lead drafting a shared AI glossary for {{team_name}}. Our product's AI surface area: {{ai_surface}}.

## Glossary structure

### Scope
- Only terms we actually use in planning, specs, or customer conversations
- Each term gets: short definition + how we use it operationally + the most common misuse
- Avoid textbook definitions; define in our product's context

### 20-30 terms to include (example categories)

**Architecture**
- Agent (in our product: X; not just "LLM-powered feature")
- RAG (what retrieval we actually do)
- Fine-tune vs. prompt vs. system prompt
- Tool use / function calling
- Orchestration / chains

**Quality and evaluation**
- Hallucination (and our specific categories)
- Eval / golden dataset
- Confidence score (what model produces vs. what we expose to users)
- Regression
- Temperature / sampling

**Cost and performance**
- Token (input vs. output)
- Latency (p50/p99)
- Throughput
- Cost per successful outcome
- Cache hit / miss

**Product surface**
- AI feature vs. AI-assisted feature vs. AI-powered
- Human-in-the-loop tiers
- Refusal / safety guardrail
- Disclosure / transparency label
- Personalization (what data we actually use)

## Review cycle
- Owned by product ops
- Reviewed quarterly
- Updates triggered by new model, new feature surface, or new customer misunderstanding

## Output
1. 20-30 term glossary with our operational definitions
2. The 5 terms most frequently misused in our planning meetings
3. A 1-page "read-me" for new joiners
4. The distribution plan (wiki, onboarding, quarterly refresh)

Customize This Prompt