How to Build Custom Research Agents in Claude Code for Enterprise Marketing Research Teams

Enterprise Marketing Research teams spend 15-20 hours per week on competitive intelligence, market sizing, and trend analysis. Generic research tools return generic insights. Claude Code research agents filter every query through your company context, ICP profiles, and strategic priorities. The output is research tailored to your specific market position, not boilerplate reports.

This guide provides the complete setup for enterprise research agents. You will install the Perplexity MCP for deep research capabilities, create context-aware agent skills, and build workflows for competitive analysis, market validation, and content gap identification. By the end, your research team will have autonomous agents running parallel research tasks while they focus on strategic analysis.

TL;DR

Claude Code research agents combine multiple MCP servers with your business context profiles. Install Perplexity for deep research reports, You.com for fast web search (93% accuracy, 466ms), Tavily for content extraction, and Microsoft 365 for internal SharePoint/Outlook/Teams data. Create context profiles for your company, ICP, competitors, and market positioning. Build specialized research agents for competitive intelligence, market sizing, and trend analysis. Run parallel research tasks across multiple MCPs while agents focus on strategic analysis.

Key Takeaways

• Four research MCPs cover enterprise needs: Perplexity for depth, You.com for speed, Tavily for extraction, Microsoft 365 for internal data
• You.com delivers 93% accuracy on SimpleQA with 466ms response times for real-time competitive monitoring
• Perplexity deep research eliminates timeout issues that break workflow automation tools
• Microsoft 365 MCP connects SharePoint, OneDrive, Outlook, and Teams for internal research
• Context profiles (company, ICP, competitors) ensure research relevance to your specific market
• Research agents reference your context automatically when conducting any analysis
• Parallel agent instances run multiple research tasks simultaneously across different MCPs
• Omnisearch MCP combines multiple providers in a single configuration for simplified management

What Is a Claude Code Research Agent and Why Do Enterprise Teams Need One?

A Claude Code research agent is an autonomous AI assistant that conducts market research filtered through your business context. Unlike generic research tools, these agents understand your company positioning, target audience, competitive landscape, and strategic priorities. Every research query returns insights relevant to your specific situation.

Enterprise Marketing Research teams manage multiple research streams: competitive intelligence, market sizing, customer insights, trend analysis, and content gap identification. Each stream requires different methodologies and outputs. A single research agent with access to your full context handles all streams while maintaining consistency.

The agent architecture uses three components: the Perplexity MCP for deep research capabilities, context profiles that define your business environment, and agent skills that specify research methodologies. When you request competitive analysis, the agent reads your competitor profiles, applies your strategic framework, and returns actionable intelligence.

Generic research tools miss context. They return the same competitive analysis for a Fortune 500 enterprise as they would for a startup. Context-aware agents produce research that accounts for your market position, resource constraints, and strategic objectives.

Action item: Identify the three research streams consuming the most analyst hours on your team. These become your first agent automation targets.

What Research MCP Servers Should Enterprise Teams Install?

Enterprise Marketing Research teams need multiple research tools for different use cases. Perplexity excels at deep research reports. You.com delivers the fastest web search with 93% accuracy on SimpleQA benchmarks. Tavily provides real-time search with content extraction. Microsoft 365 connects your internal SharePoint, OneDrive, and Outlook data. Installing all four creates a comprehensive research stack.

Research MCP Comparison Table

MCP Server	Best For	Speed	Depth	Cost
Perplexity	Deep research reports	Slow (2-5 min)	Highest	$0.05-0.15/query
You.com	Fast web search, news	Fast (466ms)	Medium	Usage-based
Tavily	Real-time search, extraction	Fast (500ms)	Medium	$0.01/query
Microsoft 365	Internal docs, email, Teams	Fast	Internal only	Free (included)

How Do You Install the Perplexity MCP Server?

The Perplexity MCP server connects Claude Code to deep research capabilities. Deep research generates multi-page reports with citations, trend analysis, and comprehensive market data. The MCP integration eliminates timeout issues that plague workflow automation tools.

Navigate to your Perplexity account settings. Scroll to API Keys section. Generate a new API key. Budget approximately $0.05-0.15 per deep research report. Enterprise teams running 50+ reports weekly should budget $200-400 monthly.

{
  "mcpServers": {
    "perplexity": {
      "command": "npx",
      "args": ["-y", "@anthropic/perplexity-mcp"],
      "env": {
        "PERPLEXITY_API_KEY": "pplx-xxxxxxxxxxxxxxxxxxxxxxxxxxxx"
      }
    }
  }
}

Perplexity provides three tools: search for quick queries, reasoning for complex analysis, and deep_research for comprehensive reports.

How Do You Install the You.com MCP Server?

You.com delivers the fastest web search API with 93% accuracy on SimpleQA benchmarks. Response time averages 466ms. The API is optimized for LLMs with long-form, citation-ready outputs. Enterprise teams use You.com for real-time competitive monitoring, news alerts, and fact-checking.

You.com offers multiple endpoints: Web Search for general queries, News API for breaking news, Deep Search for verified excerpts, and Content API for full webpage retrieval.

{
  "mcpServers": {
    "youcom": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://mcp.you.com/mcp/?youApiKey=YOUR_API_KEY"],
      "env": {}
    }
  }
}

Get your API key from the You.com developer portal. You.com is available on AWS Marketplace and Databricks MCP Marketplace for enterprise procurement.

How Do You Install the Tavily MCP Server?

Tavily provides real-time web search with content extraction capabilities. The tavily-search tool returns relevant results. The tavily-extract tool pulls clean content from URLs. Combined, they enable research workflows that search, extract, and synthesize.

{
  "mcpServers": {
    "tavily-mcp": {
      "command": "npx",
      "args": ["-y", "tavily-mcp@latest"],
      "env": {
        "TAVILY_API_KEY": "tvly-xxxxxxxxxxxxxxxxxxxx"
      }
    }
  }
}

Tavily also offers a remote MCP option requiring no local installation:

{
  "mcpServers": {
    "tavily-remote": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://mcp.tavily.com/mcp/?tavilyApiKey=YOUR_API_KEY"]
    }
  }
}

How Do You Install the Microsoft 365 MCP Server?

The Microsoft 365 MCP connects Claude Code to SharePoint, OneDrive, Outlook, and Teams. Research agents access internal competitive analysis, market research archives, and customer feedback stored in your Microsoft ecosystem.

For enterprise deployment, use the official Anthropic connector or community options:

Option 1: Softeria MS 365 MCP (Recommended)

{
  "mcpServers": {
    "ms365": {
      "command": "npx",
      "args": ["-y", "@softeria/ms-365-mcp-server", "--org-mode"]
    }
  }
}

The --org-mode flag enables Teams, SharePoint, and organization features. Without it, only personal account features are available.

Option 2: Office 365 MCP Server (Full Features)

{
  "mcpServers": {
    "office-mcp": {
      "command": "node",
      "args": ["/path/to/office-365-mcp-server/index.js"],
      "env": {
        "OFFICE_CLIENT_ID": "your-azure-app-client-id",
        "OFFICE_CLIENT_SECRET": "your-azure-app-client-secret",
        "OFFICE_TENANT_ID": "your-tenant-id",
        "SHAREPOINT_SYNC_PATH": "/path/to/sharepoint",
        "ONEDRIVE_SYNC_PATH": "/path/to/onedrive"
      }
    }
  }
}

This option requires Azure AD app registration with permissions for Mail.Read, Calendars.Read, Files.Read.All, and Sites.Read.All.

How Do You Install the Omnisearch MCP for Multiple Providers?

The mcp-omnisearch package combines multiple search providers in a single MCP server. Configure the API keys you have. Unused providers are automatically skipped.

{
  "mcpServers": {
    "mcp-omnisearch": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "mcp-omnisearch"],
      "env": {
        "TAVILY_API_KEY": "your-tavily-key",
        "PERPLEXITY_API_KEY": "your-perplexity-key",
        "BRAVE_API_KEY": "your-brave-key",
        "JINA_AI_API_KEY": "your-jina-key"
      }
    }
  }
}

Omnisearch simplifies configuration when you need multiple search providers without managing separate MCP servers.

Complete Enterprise Research MCP Configuration

Here is a complete mcp.json file with all research tools configured:

{
  "mcpServers": {
    "perplexity": {
      "command": "npx",
      "args": ["-y", "@anthropic/perplexity-mcp"],
      "env": {
        "PERPLEXITY_API_KEY": "pplx-xxxxxxxxxxxx"
      }
    },
    "youcom": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://mcp.you.com/mcp/?youApiKey=YOUR_KEY"],
      "env": {}
    },
    "tavily-mcp": {
      "command": "npx",
      "args": ["-y", "tavily-mcp@latest"],
      "env": {
        "TAVILY_API_KEY": "tvly-xxxxxxxxxxxx"
      }
    },
    "ms365": {
      "command": "npx",
      "args": ["-y", "@softeria/ms-365-mcp-server", "--org-mode"]
    }
  }
}

Verify All MCP Connections

Restart Claude Code to load the new configuration. Run /mcp to check all servers:

# Launch Claude Code
claude

# Check MCP servers
/mcp

You should see all configured servers listed as connected. Test each with a simple query before building agents that depend on them.

Action item: Install at least two research MCPs (Perplexity for depth, You.com or Tavily for speed). Verify connections before proceeding to agent setup.

How Do You Create Context Profiles for Enterprise Research?

Context profiles are JSON or markdown files that define your business environment. Research agents reference these profiles when conducting any analysis. The more detailed your profiles, the more relevant your research outputs become.

Enterprise research teams need four core profiles: Company Profile, ICP Profile, Competitor Profile, and Market Context Profile. Each profile serves a specific purpose in the research workflow.

Company Profile Template

Create a file at /context/company-profile.json:

{
  "company": {
    "name": "{{COMPANY_NAME}}",
    "industry": "{{INDUSTRY}}",
    "sub_industry": "{{SUB_INDUSTRY}}",
    "founding_year": {{YEAR}},
    "employee_count": "{{RANGE}}",
    "annual_revenue": "{{RANGE}}",
    "headquarters": "{{LOCATION}}",
    "markets_served": ["{{MARKET_1}}", "{{MARKET_2}}", "{{MARKET_3}}"]
  },
  "positioning": {
    "mission": "{{MISSION_STATEMENT}}",
    "value_proposition": "{{CORE_VALUE_PROP}}",
    "key_differentiators": [
      "{{DIFFERENTIATOR_1}}",
      "{{DIFFERENTIATOR_2}}",
      "{{DIFFERENTIATOR_3}}"
    ],
    "brand_voice": "{{VOICE_DESCRIPTION}}",
    "market_position": "{{LEADER_CHALLENGER_NICHE}}"
  },
  "products": [
    {
      "name": "{{PRODUCT_NAME}}",
      "category": "{{CATEGORY}}",
      "target_segment": "{{SEGMENT}}",
      "price_range": "{{PRICE}}",
      "key_features": ["{{FEATURE_1}}", "{{FEATURE_2}}"]
    }
  ],
  "strategic_priorities": {
    "current_year": ["{{PRIORITY_1}}", "{{PRIORITY_2}}", "{{PRIORITY_3}}"],
    "growth_targets": {
      "revenue": "{{TARGET}}",
      "market_share": "{{TARGET}}",
      "customer_acquisition": "{{TARGET}}"
    }
  },
  "constraints": {
    "budget_level": "{{HIGH_MEDIUM_LOW}}",
    "team_size": "{{SIZE}}",
    "technology_maturity": "{{LEVEL}}",
    "regulatory_considerations": ["{{REGULATION_1}}", "{{REGULATION_2}}"]
  }
}

ICP Profile Template

Create a file at /context/icp-profile.json:

{
  "ideal_customer_profile": {
    "company_characteristics": {
      "industry": ["{{INDUSTRY_1}}", "{{INDUSTRY_2}}"],
      "company_size": "{{EMPLOYEE_RANGE}}",
      "revenue_range": "{{REVENUE_RANGE}}",
      "geography": ["{{REGION_1}}", "{{REGION_2}}"],
      "technology_adoption": "{{EARLY_ADOPTER_MAINSTREAM_LAGGARD}}"
    },
    "buyer_personas": [
      {
        "title": "{{JOB_TITLE}}",
        "department": "{{DEPARTMENT}}",
        "seniority": "{{LEVEL}}",
        "responsibilities": ["{{RESP_1}}", "{{RESP_2}}"],
        "goals": ["{{GOAL_1}}", "{{GOAL_2}}"],
        "challenges": ["{{CHALLENGE_1}}", "{{CHALLENGE_2}}"],
        "success_metrics": ["{{METRIC_1}}", "{{METRIC_2}}"]
      }
    ],
    "pain_points": {
      "operational": ["{{PAIN_1}}", "{{PAIN_2}}"],
      "strategic": ["{{PAIN_3}}", "{{PAIN_4}}"],
      "financial": ["{{PAIN_5}}", "{{PAIN_6}}"]
    },
    "buying_behavior": {
      "decision_process": "{{COMMITTEE_CHAMPION_CONSENSUS}}",
      "typical_sales_cycle": "{{DURATION}}",
      "key_evaluation_criteria": ["{{CRITERIA_1}}", "{{CRITERIA_2}}"],
      "common_objections": ["{{OBJECTION_1}}", "{{OBJECTION_2}}"]
    },
    "content_preferences": {
      "formats": ["{{FORMAT_1}}", "{{FORMAT_2}}"],
      "channels": ["{{CHANNEL_1}}", "{{CHANNEL_2}}"],
      "topics_of_interest": ["{{TOPIC_1}}", "{{TOPIC_2}}"],
      "trusted_sources": ["{{SOURCE_1}}", "{{SOURCE_2}}"]
    }
  }
}

Competitor Profile Template

Create a file at /context/competitors.json:

{
  "competitive_landscape": {
    "direct_competitors": [
      {
        "name": "{{COMPETITOR_NAME}}",
        "website": "{{URL}}",
        "positioning": "{{THEIR_POSITIONING}}",
        "strengths": ["{{STRENGTH_1}}", "{{STRENGTH_2}}"],
        "weaknesses": ["{{WEAKNESS_1}}", "{{WEAKNESS_2}}"],
        "pricing_model": "{{MODEL}}",
        "price_range": "{{RANGE}}",
        "target_segment": "{{SEGMENT}}",
        "market_share_estimate": "{{PERCENTAGE}}",
        "recent_moves": ["{{MOVE_1}}", "{{MOVE_2}}"],
        "content_strategy": "{{DESCRIPTION}}",
        "technology_stack": ["{{TECH_1}}", "{{TECH_2}}"]
      }
    ],
    "indirect_competitors": [
      {
        "name": "{{COMPETITOR_NAME}}",
        "category": "{{CATEGORY}}",
        "overlap_areas": ["{{AREA_1}}", "{{AREA_2}}"],
        "threat_level": "{{HIGH_MEDIUM_LOW}}"
      }
    ],
    "emerging_threats": [
      {
        "name": "{{COMPANY_OR_TREND}}",
        "description": "{{DESCRIPTION}}",
        "timeline": "{{WHEN}}",
        "potential_impact": "{{IMPACT}}"
      }
    ]
  }
}

Market Context Profile Template

Create a file at /context/market-context.json:

{
  "market_context": {
    "market_definition": {
      "name": "{{MARKET_NAME}}",
      "tam": "{{TOTAL_ADDRESSABLE_MARKET}}",
      "sam": "{{SERVICEABLE_ADDRESSABLE_MARKET}}",
      "som": "{{SERVICEABLE_OBTAINABLE_MARKET}}",
      "growth_rate": "{{CAGR}}",
      "maturity_stage": "{{EMERGING_GROWTH_MATURE_DECLINING}}"
    },
    "trends": {
      "macro_trends": ["{{TREND_1}}", "{{TREND_2}}"],
      "technology_trends": ["{{TREND_3}}", "{{TREND_4}}"],
      "buyer_behavior_trends": ["{{TREND_5}}", "{{TREND_6}}"],
      "regulatory_trends": ["{{TREND_7}}", "{{TREND_8}}"]
    },
    "industry_events": {
      "conferences": ["{{EVENT_1}}", "{{EVENT_2}}"],
      "publications": ["{{PUB_1}}", "{{PUB_2}}"],
      "analysts": ["{{ANALYST_1}}", "{{ANALYST_2}}"],
      "influencers": ["{{INFLUENCER_1}}", "{{INFLUENCER_2}}"]
    },
    "seasonality": {
      "peak_periods": ["{{PERIOD_1}}", "{{PERIOD_2}}"],
      "budget_cycles": "{{DESCRIPTION}}",
      "content_calendar_considerations": ["{{CONSIDERATION_1}}", "{{CONSIDERATION_2}}"]
    }
  }
}

Action item: Create all four context profiles for your organization. Start with Company and ICP profiles as they provide the foundation for all research.

How Do You Build the Enterprise Research Agent?

The research agent is a markdown file that defines the agent’s purpose, capabilities, and instructions. Claude Code reads this file and invokes the agent when research tasks are requested. The agent references your context profiles automatically.

Research Agent Skill File

Create a file at /.claude/agents/researcher-agent.md:

# Enterprise Research Agent

## Purpose
Conduct market research, competitive intelligence, and trend analysis filtered through organizational context. This agent produces insights relevant to your specific market position, not generic industry reports.

## Model
claude-sonnet-4-20250514

## When to Invoke
- Market research requests
- Competitive analysis tasks
- Trend identification and monitoring
- Content gap analysis
- Market sizing and validation
- Customer insight gathering
- Industry benchmarking

## Context Integration

Before conducting any research, this agent MUST:

1. Read `/context/company-profile.json` to understand organizational positioning
2. Read `/context/icp-profile.json` to understand target audience
3. Read `/context/competitors.json` to understand competitive landscape
4. Read `/context/market-context.json` to understand market dynamics

All research outputs must be filtered through this context to ensure relevance.

## Research Methodologies

### Competitive Intelligence
- Analyze competitor positioning relative to our differentiation
- Identify gaps in competitor offerings we can exploit
- Track competitor content strategies and messaging
- Monitor pricing changes and new product launches
- Assess competitive threat levels based on our strategic priorities

### Market Validation
- Validate demand signals for our product categories
- Assess market size relative to our growth targets
- Identify underserved segments matching our ICP
- Evaluate pricing tolerance in target markets
- Confirm product-market fit indicators

### Trend Analysis
- Identify trends relevant to our strategic priorities
- Assess trend impact on our target audience
- Evaluate technology trends affecting our product roadmap
- Monitor regulatory trends in our operating markets
- Track buyer behavior shifts in our segments

### Content Gap Analysis
- Identify topics our ICP searches that lack quality coverage
- Analyze competitor content to find differentiation opportunities
- Discover questions our audience asks that no one answers well
- Map content opportunities to our expertise areas
- Prioritize gaps by search volume and strategic alignment

## Output Requirements

All research outputs must include:

1. Executive Summary (1 page maximum)
   - Key findings relevant to our strategic priorities
   - Recommended actions based on our constraints
   - Risk assessment specific to our market position

2. Detailed Findings
   - Data and evidence supporting each finding
   - Citations and sources for verification
   - Confidence levels for each insight

3. Strategic Implications
   - How findings affect our current initiatives
   - Opportunities aligned with our growth targets
   - Threats requiring immediate attention

4. Recommended Next Steps
   - Specific actions ranked by priority
   - Resource requirements for each action
   - Timeline recommendations

## Tool Usage

Select the appropriate MCP tool based on query requirements:

### Perplexity Tools
- `search`: Quick factual queries, simple lookups
- `reasoning`: Multi-step analysis, comparisons, explanations
- `deep_research`: Comprehensive reports, market sizing, competitive landscapes

### You.com Tools
- `web_search`: Fast real-time search (466ms average)
- `news_search`: Breaking news and recent articles
- `deep_search`: Verified excerpts for accuracy-critical queries

### Tavily Tools
- `tavily-search`: Real-time web search with filters
- `tavily-extract`: Content extraction from specific URLs
- `tavily-crawl`: Site mapping and comprehensive crawling

### Microsoft 365 Tools
- `search-sharepoint`: Internal document search
- `search-email`: Outlook message search
- `search-teams`: Teams chat and channel search
- `read-file`: OneDrive/SharePoint file access

### Tool Selection Matrix

| Research Type | Primary Tool | Secondary Tool | Use Case |
|---------------|--------------|----------------|----------|
| Deep research report | Perplexity deep_research | - | Market sizing, competitive landscape |
| Real-time news | You.com news_search | Tavily search | Breaking news, announcements |
| Fact verification | You.com web_search | Tavily search | Quick accuracy checks |
| Content extraction | Tavily extract | - | Pull content from specific URLs |
| Internal research | MS365 search | - | Past research, customer feedback |
| Competitive monitoring | Perplexity reasoning | You.com search | Compare multiple competitors |

Default to `deep_research` for any request involving:
- Market sizing or TAM/SAM/SOM analysis
- Competitive landscape mapping
- Trend analysis spanning multiple sources
- Content gap identification

Use You.com or Tavily for:
- Time-sensitive queries requiring speed
- Fact-checking individual claims
- News monitoring and alerts
- Real-time competitive updates

## Quality Standards

- Never return generic insights that could apply to any company
- Always tie findings back to specific elements in context profiles
- Include confidence levels for all estimates and projections
- Provide citations for factual claims
- Acknowledge limitations and gaps in available data
- Recommend follow-up research where needed

Verify Agent Installation

Restart Claude Code and run /agents to confirm the research agent is loaded:

# Restart Claude Code
claude

# Check agents
/agents

You should see the researcher-agent listed. View the agent to confirm it loaded correctly.

Action item: Create the research agent file in your Claude Code project. Verify it appears in the agents list.

What Enterprise Research Workflows Should You Automate First?

Enterprise Marketing Research teams run recurring research workflows. Automating these workflows frees analysts for strategic interpretation while agents handle data gathering. Start with high-frequency, well-defined workflows.

Workflow 1: Weekly Competitive Intelligence Report

This workflow monitors competitor activity and produces a weekly briefing for the marketing leadership team.

Prompt Template:

SYSTEM: You are the Enterprise Research Agent conducting weekly competitive intelligence.

<context>
Reference the following context profiles:
- /context/company-profile.json
- /context/competitors.json
- /context/market-context.json
</context>

Conduct a comprehensive competitive intelligence scan for the past 7 days.

For each competitor in our profile, research:
1. New product announcements or feature releases
2. Pricing changes or new packaging
3. Content published (blog posts, whitepapers, webinars)
4. Press coverage and media mentions
5. Job postings indicating strategic direction
6. Social media engagement and messaging shifts
7. Customer reviews and sentiment changes

Use Perplexity deep_research to generate a comprehensive report.

Output format:
- Executive summary (3-5 bullet points of most significant moves)
- Competitor-by-competitor breakdown
- Threat assessment with recommended responses
- Opportunities we should exploit this week

MUST filter all findings through our strategic priorities and positioning.
MUST highlight anything requiring immediate attention.

Automation Schedule: Run every Monday morning. Store reports in /knowledge/competitive-intel/ with date stamps.

Workflow 2: Quarterly Market Sizing Update

This workflow refreshes market size estimates and validates growth assumptions.

Prompt Template:

SYSTEM: You are the Enterprise Research Agent conducting quarterly market sizing.

<context>
Reference the following context profiles:
- /context/company-profile.json
- /context/market-context.json
- /context/icp-profile.json
</context>

Conduct a comprehensive market sizing update for our target markets.

Research and update:
1. Total Addressable Market (TAM) with current data
2. Serviceable Addressable Market (SAM) based on our ICP
3. Serviceable Obtainable Market (SOM) based on our competitive position
4. Growth rate projections (1-year, 3-year, 5-year)
5. Market maturity indicators
6. Segment-level sizing for our priority segments

Use Perplexity deep_research to gather comprehensive data.

Output format:
- Market size summary table with YoY changes
- Methodology and data sources used
- Confidence levels for each estimate
- Comparison to our previous quarter estimates
- Implications for our growth targets
- Recommended adjustments to our SAM/SOM definitions

MUST reconcile findings with our current growth targets.
MUST identify any market shifts affecting our strategic priorities.

Automation Schedule: Run first week of each quarter. Present findings to leadership for planning cycles.

Workflow 3: Content Gap Analysis for Editorial Planning

This workflow identifies content opportunities aligned with audience needs and competitive whitespace.

Prompt Template:

SYSTEM: You are the Enterprise Research Agent conducting content gap analysis.

<context>
Reference the following context profiles:
- /context/icp-profile.json
- /context/competitors.json
- /context/company-profile.json
</context>

Conduct a comprehensive content gap analysis for our target audience.

Research:
1. Questions our ICP asks that lack quality answers online
2. Topics competitors cover poorly or not at all
3. Emerging topics with growing search interest
4. Pain points from our ICP profile that lack educational content
5. Industry topics where we have unique expertise
6. Content formats our audience prefers but competitors underutilize

For each gap identified, provide:
- Topic description and search intent
- Estimated search volume or demand signals
- Competitor coverage assessment (none/weak/strong)
- Our credibility to address this topic
- Recommended content format
- Priority score (1-10) based on strategic alignment

Use Perplexity deep_research to analyze the content landscape.

Output format:
- Top 10 content gaps ranked by priority
- Gap analysis methodology
- Competitor content audit summary
- 90-day editorial calendar recommendations
- Resource requirements for content production

MUST align recommendations with our brand voice and positioning.
MUST prioritize gaps where we have differentiated expertise.

Automation Schedule: Run monthly before editorial planning meetings.

Workflow 4: Trend Monitoring and Early Warning

This workflow identifies emerging trends and assesses their impact on your business.

Prompt Template:

SYSTEM: You are the Enterprise Research Agent monitoring market trends.

<context>
Reference the following context profiles:
- /context/market-context.json
- /context/company-profile.json
- /context/icp-profile.json
</context>

Conduct a comprehensive trend analysis for our market.

Research:
1. Emerging technologies affecting our industry
2. Buyer behavior shifts in our target segments
3. Regulatory changes in our operating markets
4. New entrants or business models disrupting the space
5. Economic factors affecting buyer budgets
6. Talent and skill trends affecting our customers

For each trend identified, provide:
- Trend description and evidence
- Timeline (emerging/accelerating/mainstream)
- Impact assessment on our business (high/medium/low)
- Affected stakeholders (customers, competitors, partners)
- Opportunities the trend creates for us
- Threats the trend poses to our position
- Recommended response actions

Use Perplexity deep_research for comprehensive coverage.

Output format:
- Trend radar visualization (text-based)
- Deep dive on top 5 trends by impact
- Early warning signals to monitor
- Strategic implications for our roadmap
- Recommended actions by priority

MUST assess trends against our strategic priorities.
MUST identify trends requiring immediate leadership attention.

Automation Schedule: Run bi-weekly. Escalate high-impact trends immediately.

Workflow 5: Hybrid Internal-External Research

This workflow combines Microsoft 365 internal data with external research for comprehensive analysis.

Prompt Template:

SYSTEM: You are the Enterprise Research Agent conducting hybrid research.

<context>
Reference the following context profiles:
- /context/company-profile.json
- /context/competitors.json
- /context/market-context.json
</context>

Conduct a comprehensive research project combining internal and external sources.

PHASE 1: Internal Research (Microsoft 365)
1. Search SharePoint for existing research on {{TOPIC}}
2. Search Outlook for relevant stakeholder communications
3. Search Teams for discussion threads and decisions
4. Compile findings from internal sources

PHASE 2: External Research (Perplexity/You.com/Tavily)
1. Use You.com for real-time market data and news
2. Use Perplexity deep_research for comprehensive analysis
3. Use Tavily to extract content from specific competitor pages

PHASE 3: Synthesis
1. Compare internal assumptions with external findings
2. Identify gaps between internal knowledge and market reality
3. Flag contradictions requiring resolution
4. Recommend updates to internal documentation

Output format:
- Internal findings summary
- External research summary
- Gap analysis between internal and external
- Recommended actions
- Sources and citations

MUST cite both internal documents and external sources.
MUST flag any significant discrepancies between internal assumptions and external data.

Use Case: Quarterly strategy reviews where internal planning documents need validation against current market conditions.

Action item: Select one workflow to implement first. Run it manually three times to validate outputs before scheduling automation.

How Do You Run Parallel Research Tasks?

Claude Code supports multiple terminal instances. Each instance runs an independent agent. Enterprise research teams run parallel tasks to gather comprehensive data without sequential delays.

Parallel Research Architecture

Open multiple terminal windows, each running Claude Code. Assign different research tasks to each instance:

Terminal 1: Competitive Intelligence

claude
# Run competitive analysis
Use your research agent to analyze competitor X's latest product launch and its implications for our roadmap.

Terminal 2: Market Sizing

claude
# Run market sizing
Use your research agent to estimate the TAM for AI-powered marketing tools in North America.

Terminal 3: Content Analysis

claude
# Run content gap analysis
Use your research agent to identify content gaps in the marketing automation education space.

Each agent runs independently, accessing Perplexity deep research in parallel. A 30-minute research task becomes three 30-minute parallel tasks, tripling throughput without additional time.

Resource Considerations

Perplexity API rate limits apply across all instances. Monitor your API usage to avoid throttling. Enterprise teams should:

• Set up API usage alerts at 50% and 80% of budget
• Stagger deep research queries by 2-3 minutes when running 5+ parallel tasks
• Use search for quick lookups to conserve deep research quota
• Cache common queries in your knowledge folder to avoid repeat API calls

Action item: Test parallel research with two terminals. Run different research queries simultaneously to verify your setup handles concurrent requests.

How Do You Generate Branded PDF Research Reports?

Raw research outputs need formatting for executive consumption. Claude Code’s document skills transform agent outputs into professional PDF reports with executive summaries, branded headers, and proper citations.

PDF Document Skill Setup

Claude’s official skills library includes document creation capabilities. Install the skills library if you have not already:

# In Claude Code
/install-skills

Select the document skills package. This enables Claude to create PDFs, Word documents, and formatted reports.

Report Generation Prompt

After research completes, request a formatted report:

Create a professional PDF research report with the following structure:

1. Cover page with:
   - Report title
   - Prepared for: {{COMPANY_NAME}}
   - Prepared by: Enterprise Research Agent
   - Date: {{DATE}}

2. Executive Summary (1 page)
   - 5-7 key findings tailored to our strategic priorities
   - Recommended actions with urgency levels
   - One-paragraph strategic implication

3. Full Research Report
   - Include all findings from the research
   - Maintain citations and sources
   - Add confidence levels for estimates

4. Appendix
   - Methodology description
   - Data sources list
   - Limitations and caveats

Use our brand colors: {{PRIMARY_COLOR}} for headers, {{SECONDARY_COLOR}} for accents.

Save to /knowledge/reports/{{REPORT_NAME}}-{{DATE}}.pdf

Report Template Skill

Create a reusable report template at /.claude/skills/research-report-template.md:

# Research Report Template Skill

## Purpose
Generate consistently formatted research reports for enterprise consumption.

## When to Use
- After any research agent completes a task
- When user requests formatted output
- For executive presentations and briefings

## Report Structure

### Page 1: Cover
- Report title (centered, 24pt, brand primary color)
- Subtitle with research type
- Prepared for: [Company Name]
- Date: [Report Date]
- Classification: [Internal/Confidential/Public]

### Page 2: Executive Summary
- Maximum 1 page
- Key Findings: 5-7 bullet points
- Strategic Implications: 2-3 sentences
- Recommended Actions: Prioritized list with owners
- Risk Assessment: High/Medium/Low with rationale

### Pages 3+: Detailed Findings
- Section headers matching research methodology
- Data tables where applicable
- Charts and visualizations (text-based)
- Citations in footnotes
- Confidence indicators: High (>80%), Medium (50-80%), Low (<50%)

### Final Page: Appendix
- Methodology description
- Data sources with access dates
- Limitations and assumptions
- Glossary of terms
- Contact for questions

## Formatting Standards

### Typography
- Headers: Bold, brand primary color
- Body: 11pt, black
- Citations: 9pt, gray
- Page numbers: Bottom center

### Brand Colors
Reference /context/company-profile.json for brand colors.
Default if not specified:
- Primary: #2563EB
- Secondary: #1E40AF
- Accent: #F59E0B

### File Naming
Format: [report-type]-[topic]-[YYYY-MM-DD].pdf
Example: competitive-intel-q1-review-2026-01-07.pdf

### Storage Location
Save all reports to: /knowledge/reports/
Create subdirectories by report type:
- /knowledge/reports/competitive/
- /knowledge/reports/market-sizing/
- /knowledge/reports/trends/
- /knowledge/reports/content-gaps/

Action item: Generate your first PDF report from a completed research task. Review formatting and adjust the template skill as needed.

How Do You Build Specialized Research Agents for Different Functions?

Enterprise research teams support multiple functions: product marketing, demand generation, competitive intelligence, and customer insights. Specialized agents optimize for each function’s unique requirements.

Competitive Intelligence Agent

Create /.claude/agents/competitive-intel-agent.md:

# Competitive Intelligence Agent

## Purpose
Monitor and analyze competitor activities to inform strategic decisions.

## Model
claude-sonnet-4-20250514

## Specialization
This agent focuses exclusively on competitive intelligence. It tracks competitor movements, assesses threats, and identifies opportunities.

## Context Integration
MUST read /context/competitors.json before any analysis.
MUST filter all findings through our positioning in /context/company-profile.json.

## Primary Tasks

### Win/Loss Analysis
- Research why deals were won or lost against specific competitors
- Identify patterns in competitive wins
- Recommend battlecard updates

### Competitor Monitoring
- Track product launches and feature releases
- Monitor pricing and packaging changes
- Analyze hiring patterns and job postings
- Follow press coverage and analyst mentions

### Threat Assessment
- Evaluate new market entrants
- Assess partnership announcements
- Identify potential acquisition targets
- Monitor patent filings and R&D signals

### Battlecard Updates
- Recommend messaging updates based on competitor moves
- Identify new objection handling needs
- Suggest proof points to counter competitor claims

## Output Format
All outputs must include:
- Threat level (Critical/High/Medium/Low)
- Recommended response timeline (Immediate/This Quarter/Monitor)
- Specific action items for sales and marketing
- Evidence and citations

## Tool Preferences
- Use `deep_research` for new competitor analysis
- Use `reasoning` for comparing competitor strategies
- Use `search` for quick fact-checking

Customer Insights Agent

Create /.claude/agents/customer-insights-agent.md:

# Customer Insights Agent

## Purpose
Gather and analyze customer intelligence to inform product and marketing decisions.

## Model
claude-sonnet-4-20250514

## Specialization
This agent focuses on understanding customer needs, behaviors, and preferences.

## Context Integration
MUST read /context/icp-profile.json before any analysis.
MUST consider buyer personas when interpreting findings.

## Primary Tasks

### Voice of Customer Research
- Analyze review sites for customer sentiment
- Monitor social media conversations
- Track community forum discussions
- Identify common praise and complaints

### Buyer Journey Mapping
- Research how buyers discover solutions
- Identify information sources they trust
- Map decision-making processes
- Understand evaluation criteria by persona

### Need State Analysis
- Identify emerging customer pain points
- Track changing priorities and budgets
- Assess technology adoption readiness
- Monitor regulatory concerns affecting buyers

### Segmentation Validation
- Validate ICP assumptions with market data
- Identify underserved segments
- Assess segment growth trajectories
- Recommend segment prioritization

## Output Format
All outputs must include:
- Customer evidence (quotes, data points)
- Segment specificity (which personas affected)
- Confidence level based on sample size
- Recommended marketing implications
- Suggested follow-up research

## Tool Preferences
- Use `deep_research` for comprehensive voice of customer analysis
- Use `reasoning` for segment comparison and prioritization
- Use `search` for specific customer data points

Market Sizing Agent

Create /.claude/agents/market-sizing-agent.md:

# Market Sizing Agent

## Purpose
Estimate market sizes, growth rates, and segment opportunities.

## Model
claude-sonnet-4-20250514

## Specialization
This agent focuses on quantitative market analysis and sizing.

## Context Integration
MUST read /context/market-context.json for baseline estimates.
MUST read /context/company-profile.json for growth target alignment.

## Primary Tasks

### TAM/SAM/SOM Analysis
- Calculate Total Addressable Market from multiple sources
- Define Serviceable Addressable Market based on ICP
- Estimate Serviceable Obtainable Market based on competitive position
- Triangulate estimates using multiple methodologies

### Growth Rate Estimation
- Analyze historical growth patterns
- Assess growth drivers and inhibitors
- Model scenarios (base, optimistic, pessimistic)
- Compare analyst projections

### Segment Sizing
- Size individual market segments
- Assess segment growth differentials
- Identify fastest-growing segments
- Map segment sizes to our capabilities

### Opportunity Assessment
- Quantify whitespace opportunities
- Estimate expansion potential in existing segments
- Size adjacent market opportunities
- Calculate share gain potential by competitor

## Output Format
All outputs must include:
- Sizing methodology used
- Data sources with reliability assessment
- Confidence intervals (not point estimates)
- Comparison to previous estimates
- Implications for planning and targets

## Tool Preferences
- Use `deep_research` for comprehensive market sizing
- Use `reasoning` for methodology comparison
- Use `search` for specific data points and statistics

Action item: Create specialized agents for your two highest-priority research functions. Test each with function-specific queries.

What Common Mistakes Break Enterprise Research Agents?

Research agent implementations fail for predictable reasons. Avoid these mistakes to ensure reliable outputs.

Mistake 1: Incomplete Context Profiles

Agents without complete context profiles return generic insights. A competitor profile missing recent product launches leads to outdated competitive analysis. Review and update context profiles monthly.

Mistake 2: Overly Broad Research Queries

Queries like “tell me about the market” return unfocused reports. Specific queries like “estimate the TAM for AI-powered email personalization tools in North America for mid-market B2B companies” return actionable data.

Mistake 3: Ignoring Confidence Levels

Deep research returns estimates with varying reliability. Treating all estimates equally leads to bad decisions. Always review confidence levels and pursue follow-up research for low-confidence findings.

Mistake 4: No Validation Loop

Research agents occasionally hallucinate or misinterpret sources. Critical findings require human validation. Establish review protocols for high-stakes research before acting on recommendations.

Mistake 5: Exceeding API Budgets

Deep research queries are expensive. Running unlimited parallel tasks without budget monitoring leads to surprise bills. Set up usage alerts and establish query approval workflows for expensive research.

Action item: Audit your context profiles for completeness. Update any profiles more than 60 days old.

Final Takeaways

Four research MCPs create a comprehensive enterprise stack: Perplexity for deep reports, You.com for speed and accuracy, Tavily for content extraction, and Microsoft 365 for internal data.

Context profiles for company, ICP, competitors, and market define the lens through which all research is filtered. More context produces more relevant outputs.

You.com and Tavily provide sub-500ms response times for real-time competitive monitoring and news alerts. Use them for speed-critical queries.

Microsoft 365 MCP connects your internal knowledge base to research workflows. Combine internal assumptions with external validation.

Specialized agents for competitive intelligence, customer insights, and market sizing optimize outputs for each function.