AI & Agents

LLMs, agents, and AI applied to analytics and data work.

LLM basics

• Models — GPT-4, Claude, Llama, etc.; choose by cost, latency, and capability
• Prompts — Instructions and context; structure matters
• Token limits — Input and output; affects context and cost
• Temperature — Lower = more deterministic; higher = more varied

Agents

Agents use LLMs to plan, use tools, and iterate:

• ReAct — Reason and act in a loop
• Tool use — Call APIs, run code, query data
• Orchestration — LangChain, LlamaIndex, or custom loops

Analytics use cases

• Natural language to SQL — Turn questions into queries
• Report summarization — Summarize dashboards and tables
• Data quality — Suggest checks and anomaly detection
• Documentation — Generate or update docs from code and schemas