Draft a shared AI glossary for your product team

Team-Specific AI Glossaries Beat Textbook Definitions

AI vocabulary has exploded and the same word means different things in different teams — "agent" can mean a chatbot, a tool-using LLM, a workflow orchestrator, or a fully autonomous task-completer. Textbook definitions don't solve this; a shared glossary that names what your product specifically means by each term does. The output is a 1-page artifact every new joiner reads and every planning conversation references.

How the Draft a shared AI glossary Prompt Works

The prompt scopes the glossary to 20-30 terms the team actually uses in planning, specs, and customer conversations — not a comprehensive academic vocabulary. Each term gets three things: a short definition, an operational explanation of how the team uses it in their product, and the most common misuse so new joiners learn both the right and wrong mental models.

The categories cover architecture (agents, RAG, fine-tune vs. prompt), quality and evaluation (hallucination, eval, regression), cost and performance (tokens, latency, cost per outcome), and product surface (human-in-the-loop tiers, disclosure labels). The quarterly review cycle keeps the glossary fresh as the team ships new AI surfaces and encounters new customer misunderstandings.

When to Use It

• Planning meetings stall because team members mean different things by the same word.
• A new AI feature is entering discovery and the team lacks shared vocabulary.
• New hires are spending weeks catching up on team-specific AI terminology.
• Customer-facing docs use AI terms inconsistently across marketing, support, and product.
• A company is scaling its AI surface area and needs a shared operating language.

Common Pitfalls

• Copying a textbook glossary. Textbook definitions are not team definitions. The operational "how we use it" column is the load-bearing one.
• Glossary with no owner. Without an owner, the glossary decays within a quarter. Product ops should own it.
• No distribution plan. A glossary in a wiki folder nobody reads does nothing. Include it in onboarding and meeting references.

Sources

• Anthropic Research
• AI Adoption in Product Orgs — Reforge
• The Product Engineer Role — PostHog
• GitHub Developer Research