AI Jargon Decoded: Ten Terms You Will Hear in a Sales Pitch

Hardly any conversation about AI happens without technical jargon. That is not an issue, as long as you know the key questions behind each term. Here are ten terms you will encounter when vetting vendors, along with the precise questions you should ask.

1. Large Language Model (LLM)

The AI system that understands and generates natural language. Your question: Which model is being used, and where is it hosted?

2. AI Agent

An AI that acts using tools rather than just generating text responses. Your question: Which tools will the agent access in our operations, and what actions can it take autonomously?

3. Prompt

The instruction or input given to the AI. Your question: Do our employees need specialized prompt engineering skills, or is standard business language sufficient?

4. Hallucination

A convincingly phrased, yet factually incorrect response. Your question: Where does the AI source its facts, and how can we verify the basis of its answers?

5. RAG (Retrieval-Augmented Generation)

The method where the AI retrieves information from your internal data before answering. Your question: Which of our data sources are searched, and how are user access permissions enforced?

6. Training

The process of teaching a model using data. Your question: Will our company or customer data be used to train the model? The answer should always be no.

7. Autonomy

The degree of independence the AI has to execute tasks. Your question: Can we configure autonomy levels per task and revoke permissions at any time?

8. Audit Trail (Log)

The comprehensive record tracking every action the AI has taken. Your question: Can we trace and review every automated decision and step?

9. Data Processing Agreement (DPA)

The mandatory contract pursuant to Article 28 GDPR. Your question: Is a compliant DPA provided, and which sub-processors are involved?

10. EU AI Act

The European Union's regulatory framework for AI. Your question: How does the vendor support our compliance regarding AI literacy (Article 4) and transparency requirements (Article 50)?