How do you implement AI agents for research and information synthesis?

Research agents autonomously gather, analyze, and synthesize information from multiple sources. Architecture: query planner (decompose research question into sub-questions and search strategies), information gatherer (execute web searches, read documents, query databases), content analyzer (extract key findings, assess source credibility, identify contradictions), and synthesizer (compile findings into coherent, well-sourced reports). Implementation: the agent receives a research topic, generates a research plan (key questions, search queries, sources to consult), executes searches in parallel, evaluates source credibility (domain authority, publication date, author expertise), extracts relevant information with citations, identifies knowledge gaps and conducts follow-up searches, and synthesizes findings into a structured report. Tool suite: web search (multiple search engines for diverse results), web scraper (extract content from search results), document reader (process PDFs, academic papers), calculator (numerical analysis), and report writer (format findings with citations). Quality measures: source triangulation (verify claims across multiple independent sources), recency weighting (prefer recent sources for time-sensitive topics), expertise filtering (prioritize authoritative sources), and contradiction flagging (highlight when sources disagree). Output format: executive summary, detailed findings organized by subtopic, source bibliography, confidence assessment per finding, and identified knowledge gaps. Advanced: implement iterative deepening (start broad, drill into promising areas), cross-reference findings with known facts, and maintain a research graph linking findings to sources and sub-questions. Evaluation: assess completeness (did the agent cover all relevant aspects), accuracy (are findings correct and well-sourced), and insight quality (did the synthesis provide useful analysis beyond raw facts).