Sign In
Signed out 5
Module 07 · Tier 2 — Basic Prompting

AI for Research & Analysis

Surface insights fast. Know the difference between AI-assisted and AI-generated research.

11 min Tier 2 — Basic Prompting
Why This Matters

AI can surface insights that would take hours to find manually. It can synthesize large documents, identify patterns across multiple sources, and generate frameworks for thinking about complex problems. It is not, however, a research database. Understanding the difference — and knowing which uses are high-value versus high-risk — will make you dramatically more effective at research tasks while keeping you out of trouble.

The Concept

What AI Can and Cannot Do for Research

High-value AI research uses:

  • Synthesizing documents you provide — paste a 50-page report and ask for the key findings
  • Explaining complex topics at the right level — "explain GDPR implications as if to a non-legal business manager"
  • Generating frameworks — "what are the main dimensions I should evaluate when assessing a new vendor?"
  • Identifying gaps in your research — "given what I've shared, what questions am I not asking that I should be?"
  • Comparing perspectives — "summarize the main arguments for and against [position]"
  • Finding connections — "what do these three trends have in common that isn't immediately obvious?"

High-risk AI research uses (verify everything):

  • Asking for specific statistics, dates, or data points — these are prime hallucination territory
  • Asking for citations — AI will generate plausible-looking citations that often don't exist or are slightly wrong
  • Asking about recent events — the model's training cutoff means anything recent is unreliable
  • Asking for specific claims about real people or organizations

The Document Analysis Workflow

One of the most practical and reliable AI research uses: document analysis. The pattern is simple — you provide the source material, AI analyzes it. You've eliminated hallucination risk for any claim directly derived from the document.

High-value document analysis prompts:

  • "What are the three most important findings in this document, stated as actionable insights?"
  • "What assumptions does this document make that are never explicitly stated?"
  • "Summarize this for an executive who has five minutes and needs to know whether to approve this project."
  • "What questions does this document raise that it doesn't answer?"
  • "Compare these two documents: where do they agree, and where do they contradict each other?"

The Critical Evaluation Habit

Good researchers apply critical evaluation to everything they read. AI-generated research summaries require the same treatment — and in some ways more, because AI presents everything with the same confident prose style regardless of how certain or uncertain the underlying claim is.

When reviewing AI research output, ask:

  • What is the source of each specific claim? Is it from a document I provided, or from training data?
  • Are any statistics or citations here worth verifying against primary sources?
  • What perspective is this analysis missing? What has AI not included?
  • Does this synthesis represent the full picture, or did AI over-index on certain sources or framings?
Document analysis in practice

The most reliable research workflow: bring your own documents.

If you have a long report to analyze, paste the full text (or use a tool that supports file upload) and try this sequence:

  1. "Give me a 3-sentence executive summary of this document."
  2. "What is the document's central argument or recommendation?"
  3. "What evidence does it provide for that argument? How strong is the evidence?"
  4. "What is this document missing or assuming that isn't justified by the evidence?"

This sequence moves from summary → argument → evidence → critique. By the end, you understand the document better than most people who read it fully — and you have critical questions to dig into.

Hands-On Exercise

Analyze a document with structured prompts

ClaudeChatGPTGemini
Find a document you need to understand for your work — a report, a long email thread, a policy document, an article. Paste it into an AI tool (or use the file upload if available) and run through this analysis sequence: 1. Executive summary (3 sentences) 2. Main argument or recommendation 3. Key evidence cited 4. What's missing or assumed 5. Your top two follow-up questions After the AI gives you its analysis: do you agree? What did it get right? What did it miss? What would you add?
The last step — your own evaluation of AI's analysis — is the most important. This is where you develop critical judgment about AI-generated research.
Active Recall

Before moving on — close this lesson and answer these from memory. Then come back and check. Testing yourself (not re-reading) is how this sticks.

1 What is the key difference in reliability between "analyzing documents you provide" and "asking AI for research facts"?
2 Name three types of AI research outputs that should always be verified against primary sources before use.
Reflection

Where do you currently do the most research in your work? Which parts of that research process would be genuinely faster and better with AI assistance? Which parts would you be more careful about?

Key Takeaway

AI is powerful for synthesis, framework generation, and document analysis — especially when you provide the source material. It is unreliable for specific facts, citations, and recent events. Bring your own documents and verify any specific claims before using them.