AI Automation / AI Agents
Create an agent that searches, summarizes, and synthesizes information from multiple sources.
Why This Prompt Exists
Research is time-consuming: searching, reading, extracting, synthesizing. An AI agent can do all of this — but needs the right workflow.
You get:
- agents that search too narrowly (miss important sources)
- agents that search too broadly (drown in irrelevant results)
- summaries that miss key findings (shallow extraction)
- no source prioritization (low-quality sources weighted equally)
- no synthesis across sources (list of summaries, not integrated insights)
But research agents need structure:
- query generation: convert research question into search queries
- source selection: prioritize authoritative sources
- extraction: pull key claims, statistics, and quotes
- synthesis: integrate findings across sources
- citation: track sources for verifiability
Without workflow, research agents produce garbage.
This prompt designs effective research agent workflows.
The Prompt
Assume the role of a research automation architect who designs AI research agents.
Your task is to create a workflow for an agent that performs research tasks.
Generate:
1. RESEARCH QUESTION
- What the agent needs to answer
2. SEARCH STRATEGY
- Query generation: [list of search queries to use]
- Sources to search: [e.g., Google, academic databases, news, internal docs]
- Date range: [e.g., "last 12 months"]
- Number of results per source: [X]
3. SOURCE PRIORITIZATION
- Tier 1 (highest authority): [e.g., peer-reviewed journals, official reports]
- Tier 2 (moderate authority): [e.g., industry blogs, news articles]
- Tier 3 (low authority): [e.g., social media, forums]
4. EXTRACTION PROTOCOL
- For each source, extract:
* Main claim
* Key evidence (statistics, quotes)
* Methodology (if research study)
* Limitations (if acknowledged)
5. SYNTHESIS APPROACH
- Theme clustering: group similar findings
- Agreement: what do sources agree on?
- Disagreement: where do sources conflict?
- Gap identification: what's missing from the literature?
6. OUTPUT FORMAT
- Executive summary (1 paragraph)
- Key findings (bullet points with citations)
- Conflicting evidence (if any)
- Research gaps
- Source list with authority ratings
7. READY-TO-USE AGENT PROMPT
- The system prompt for the research agent
INPUTS:
Research question:
[E.G., "What are the latest trends in AI-powered customer support?"]
Sources available:
[E.G., "Google Search, internal knowledge base, competitor websites"]
Depth required:
[QUICK (10 sources, 1 hour) / STANDARD (30 sources, 1 day) / DEEP (100+ sources, 1 week)]
Output audience:
[EXECUTIVE / ANALYST / RESEARCHER]
RULES:
- Prioritize recent sources for fast-moving topics
- Include authority tiers in output (so users know which sources to trust)
- Always cite sources (uncited claims are useless for research)
- Flag limitations and conflicts — don't smooth over disagreements
- For deep research, include a verification step (check critical claims)
- Respect copyright and terms of service for each source
How To Use It
- Start with a specific, well-defined research question (vague questions produce vague results).
- Prioritize source authority — a low-authority source can still be useful, but flag it.
- Always include citations — research without citations is opinion.
- Flag conflicts and gaps explicitly — don’t smooth over disagreements.
- For high-stakes research, have a human verify critical claims.
Example Input
Research question:
“What are the latest trends in AI-powered customer support?”
Sources available:
“Google Search, Gartner reports, competitor websites”
Depth required:
“STANDARD”
Output audience:
“EXECUTIVE”
Why It Works
Most research agents are just search + summarize — missing the critical synthesis step that turns information into insight.
This framework improves outcomes by forcing:
- search strategy (what queries, what sources?)
- source prioritization (which sources to trust more?)
- extraction protocol (what to pull from each source?)
- synthesis approach (how to integrate across sources?)
- output formatting (executive summary, key findings, citations)
Great research agents don’t just find information — they synthesize it into actionable insights with clear provenance.
Build Better AI Systems
Subscribe for advanced prompt engineering, AI coding tools, debugging frameworks, and practical strategies for developers and engineers.