Research Workflow

Claude can help with research synthesis, landscape analysis, and structured summarisation of documents you provide. A disciplined research workflow — defining questions precisely, decomposing into sub-questions, synthesising across answers — produces more useful outputs than open-ended queries.

Defining the Research Question Precisely

Vague research questions produce vague answers. Before starting, refine your question to be specific about scope, depth, and output format:

Ask Claude to help sharpen a vague question: "I want to research [topic] for the purpose of [goal]. What specific questions should I be asking to make this research useful?"

Breaking Research into Sub-Questions

Complex research topics are better handled as a sequence of sub-questions rather than one large query. This keeps each answer focused and lets you build up understanding incrementally:

1. Define the landscape: "What are the main approaches to [topic]? Give me 4–6 categories with one-line descriptions."
2. Deep dive each category: "Tell me more about [category]. What are the main vendors/tools/methods? What are the tradeoffs?"
3. Comparative analysis: "Comparing [option A] and [option B] for [my specific use case], what are the key differences?"
4. Edge cases and caveats: "What are the failure modes or known limitations of [option]?"
5. Practical implications: "Given [my context], which approach would you recommend and why?"

Synthesis: Combining Answers into a Unified View

After gathering answers to multiple sub-questions, synthesise them into a coherent picture:

• "Based on our conversation so far, what is the single most important consideration when choosing [X] for [use case]?"
• "Summarise the key tradeoffs across all the options we discussed as a comparison table with columns: [option], [criteria 1], [criteria 2], [criteria 3]."
• "What are the three things I now know that I didn't know before this research?"
• "What open questions remain that this research hasn't resolved?"

Structuring Findings into a Brief or Report

Once the research is complete, ask Claude to structure the findings into a usable output:

• Research brief (1 page): "Write a one-page research brief summarising what we found. Structure: Executive Summary (3 bullets), Key Findings (4–6 points), Recommended Action, Open Questions."
• Comparison table: "Produce a comparison table of the 4 options we discussed, with one row per option and columns for [criteria 1–5]."
• Decision framework: "Based on this research, create a decision framework — a set of questions someone should answer to decide which option fits their situation."

Validating Key Claims with External Sources

Claude synthesises from its training data, which has a knowledge cutoff and may contain errors. For research that will inform decisions:

• Flag for verification: Ask Claude: "Which specific claims in this research should I verify from primary sources before acting on them?" Claude will typically flag statistics, recent events, and version-specific technical details.
• Use Claude with web search (claude.ai): Claude.ai can search the web to supplement training data with current information — useful for fast-moving topics like AI tooling, regulation, and pricing.
• Provide primary sources yourself: Upload the actual documentation, papers, or reports you want Claude to synthesise from. This is more reliable than asking Claude to recall from training data for technical detail.

Checklist: Do You Understand This?

• Start with a precise research question — scope, depth, and output format defined before the first prompt
• Decompose complex topics into 4–5 sequential sub-questions; build understanding incrementally
• Synthesise via explicit summary requests: comparison tables, decision frameworks, research briefs
• Ask Claude which claims to verify externally before acting on them
• Upload primary sources for technical depth — more reliable than relying on training data for precise specifications