AI Terms That Help Product Teams Work Together

AI terms can be hard to understand when a team discusses a product. NN/G provides a glossary to explain words used in AI and UX work. It is a reference, so readers can find the term they need. The definitions follow product use, but other people may use words differently. Knowing the differences helps teams discuss features and limits more clearly.

One important difference is between a model and the product around it. A model is only one part of an AI service. The product also provides instructions, information, tools, memory, and ways to protect users. These parts affect what people can actually do with the service. A model's ability alone does not describe the whole user experience.

The glossary also explains the difference between context and memory in AI. Context is the information available to a model for one request. It can include instructions, earlier messages, documents, and results from tools. Memory is information that the product keeps for later use. The product can bring that saved information into a future request. A context window limits how much information the model can consider at once.

An agent can use a model and tools to work toward a goal. It may choose several steps without receiving a new instruction for each one. But agents do not all have the same freedom to act. The surrounding product sets permissions, stopping rules, and opportunities for human review. Calling something an agent does not explain its exact level of control.

Teams also need ways to check behavior instead of trusting every answer. Evals are repeatable tests with clear rules for judging a result. Grounding connects answers to chosen sources, but those sources can still contain errors. Human review only helps when people can understand a result and change it. These terms describe different parts of making an AI product useful and reliable.