Synonyms: generative AI, conversational AI, generative language model, genAI model, neural language model
Do not index
Definition
An LLM is an AI model trained on large amounts of text data to generate and understand human language.
It can analyze prompts, predict likely responses, summarize information, answer questions, and generate content inside a user interface.
Use cases
Designers who treat LLMs like predictable software often ship unreliable experiences. Users lose trust quickly when the model gives incorrect or inconsistent answers.
A PM treats an AI summarizer like a search engine and removes source citations to simplify the UI. Users spot inconsistencies they can’t verify, and trust collapses within days.
A designer drops a chatbot into a checkout flow assuming it will follow fixed business rules. The model hallucinates return policies during edge-case questions.
How it's used in practice
Prompt engineering: Collaborating with developers to design precise instructions (prompts) that guide the LLM to generate user-centric and accurate responses.
Error handling: Designing user-friendly states for when the LLM hallucinates (generates incorrect info) or fails to understand a request.
Content strategy: Using LLMs to generate onboarding copy, support responses, or dynamic UI text.
🪄
Pro-tip: Remember that an LLM is a tool. Designers are still responsible for the final user experience. Always review and edit LLM-generated content to ensure it aligns with your brand voice and user needs.
Challenges & limitations
Bias: They can inherit biases from their training data, potentially leading to unfair or harmful outputs.
Integration: Seamlessly integrating an LLM into an existing product while maintaining a fast and reliable user experience can be challenging.
_.jpeg?table=block&id=3792a971-edb7-80da-a5e7-dbb82fea7984&cache=v2)
