What you’ll build
You’ll build an intelligent sales preparation assistant that researches companies, contacts, and opportunities in parallel, then generates comprehensive meeting prep documents. When a sales rep says “Prepare me for a meeting with Acme Corp,” the workflow automatically researches the company, key contacts, recent news, and potential opportunities, delivering a structured brief within seconds. This workflow demonstrates how to:- Automate pre-meeting research that typically takes 30-60 minutes
- Use parallel agents to gather information from multiple sources simultaneously
- Integrate web search and CRM data for comprehensive intelligence
- Generate structured, actionable meeting prep documents
- Trigger complex workflows from natural language chat
What you’ll build
Your sales team will be able to:- Chat: “Prepare me for a meeting with [Company Name]”
- Get a complete meeting brief including:
- Company overview and recent news
- Key contact profiles and backgrounds
- Open opportunities and deal status
- Talking points and questions to ask
- Competitor intelligence
- Risk factors and concerns
Prerequisites
Before you begin, ensure you have:- MagOneAI instance with workflow builder and chat interface access
- LLM provider configured (GPT-4, Claude 3.5 Sonnet recommended)
- Web search tool configured (Google Search API, Bing Search, or SerpAPI)
- CRM integration (optional but recommended) - Salesforce, HubSpot, or similar
- Sales intelligence tools (optional) - LinkedIn Sales Navigator, ZoomInfo, etc.
Architecture
The sales intelligence workflow uses parallel research agents to gather comprehensive information quickly:Why this architecture works
Speed
Four parallel research agents complete in ~15-20 seconds vs. 2+ minutes sequential
Comprehensive
Multi-source research provides complete context: public data, CRM history, and real-time news
Natural Language
Sales reps use natural chat commands - no forms or complex interfaces
Actionable Output
Structured briefs include talking points, questions, and next steps - not just raw data
Step-by-step build
Create the research agents
You’ll create five agents: four research specialists and one synthesis agent.
1. Company Research Agent
Name: Company Intelligence Researcher Model: GPT-4 or Claude 3.5 Sonnet Tools: Web SearchPersona:2. Contact Research Agent
Name: Contact Intelligence Researcher Model: GPT-4 or Claude 3.5 Sonnet Tools: Web Search, LinkedIn (if integrated)Persona:3. Opportunity Analysis Agent
Name: Deal Intelligence Analyst Model: GPT-4 or Claude 3.5 Sonnet Tools: CRM API (if integrated)Persona:4. News & Updates Agent
Name: News Intelligence Monitor Model: GPT-4 or Claude 3.5 Sonnet Tools: News Search API, Web SearchPersona:5. Synthesis Agent
Name: Meeting Prep Coordinator Model: GPT-4 or Claude 3.5 SonnetPersona:Configure tools and integrations
Set up the external tools your agents will use.
Web Search Tool
- Navigate to Tools in your MagOneAI project
- Add Web Search Tool
- Configure with your search API:
- Google Custom Search API
- Bing Search API
- SerpAPI
- Or use MagOneAI’s built-in search if available
- Test the search tool with a sample query
CRM Integration (Optional)
- Add CRM Tool or HTTP API Tool
- Configure authentication for your CRM:
- Salesforce: OAuth 2.0
- HubSpot: API key
- Other: API credentials
- Create API endpoints for:
- Search accounts by company name
- Get open opportunities for account
- Get contact history
- Test the integration
If CRM integration isn’t available initially, the workflow still provides value through web and news research. You can add CRM later without rebuilding the workflow.
News Search (Optional)
- Add News API Tool or use web search with news filters
- Configure with news API credentials if using dedicated service
- Set search parameters to prioritize recent news (last 30 days)
Build the workflow
Construct the workflow in the MagOneAI workflow builder.
-
Add Chat Trigger
- Type: Chat Trigger
- Configure trigger patterns:
- “Prepare me for a meeting with [company]”
- “Meeting prep for [company]”
- “Research [company] for meeting”
- “Brief me on [company]”
- Extract
company_namefrom user input
-
Add Company Name Extraction (Optional)
- If your platform supports it, add a simple agent to parse company name
- Or use regex extraction in trigger configuration
- Store in variable:
company_name
-
Add Parallel Research Node
- Drag a Parallel Node onto the canvas
- Connect to trigger
- Configure four parallel branches
Configure parallel research branches
Set up four research branches that execute simultaneously.Branch 1 - Company Research
- Agent Node: Company Intelligence Researcher
- Input:
{{trigger.company_name}} - Enable tools: Web Search
- Timeout: 30 seconds
- Agent Node: Contact Intelligence Researcher
- Input:
{{trigger.company_name}} - Enable tools: Web Search, LinkedIn API (if configured)
- Timeout: 30 seconds
- Agent Node: Deal Intelligence Analyst
- Input:
{{trigger.company_name}} - Enable tools: CRM API (if configured)
- Timeout: 20 seconds
- Handle errors gracefully: If no CRM data, return empty state
- Agent Node: News Intelligence Monitor
- Input:
{{trigger.company_name}} - Enable tools: News Search API or Web Search
- Timeout: 30 seconds
Add synthesis agent
After parallel research completes, synthesize results into meeting prep document.
- Add Agent Node: Meeting Prep Coordinator
- Configure inputs:
- Output format: Markdown
- Timeout: 30 seconds
Output to chat
Display the meeting prep document in the chat interface.
- Add Output Node or let the workflow auto-return to chat
- Format output:
- Add option to save: Include buttons like “Save to CRM” or “Email to me”
Consider adding a final step that automatically creates a note in your CRM with the meeting prep document for future reference.
Test the workflow
Test with various company names and scenarios.Test Case 1: Well-known public company
- Command: “Prepare me for a meeting with Microsoft”
- Expect: Rich public information, recent news, executive profiles
- Verify: All research branches complete successfully
- Command: “Meeting prep for [local business]”
- Expect: Less public data, still useful overview and news
- Verify: Agents handle limited data gracefully
- Command: “Brief me on [existing customer]”
- Expect: CRM data enriches the brief with deal history
- Verify: Opportunity analysis includes actual CRM data
- Command: “Research Amazn for meeting” (typo)
- Verify: Agents interpret correctly or ask for clarification
- Parallel execution timing (should be ~15-30s total)
- Agent output quality and relevance
- Proper error handling when data isn’t available
- Markdown formatting in chat output
Add enhancements
Improve the workflow with additional features.Save to CRM
- Add a button: “Save this brief to CRM”
- Create a CRM note/activity with the meeting prep content
- Link to the company record
- Add a button: “Email this to me”
- Format as HTML email
- Include attachments (if any news articles or reports)
- Add button: “Set follow-up reminder”
- Create calendar event or task
- Include key talking points in event description
- Add optional 5th parallel branch
- Research mentioned competitors
- Compare positioning and recent news
Key concepts demonstrated
Chat-Initiated Workflows
Trigger complex multi-agent workflows from natural language chat commands
Web Search Integration
Use search APIs to gather real-time public information from across the web
Parallel Research
Execute multiple research agents simultaneously for 4x faster results
CRM Integration
Enrich AI research with structured data from your CRM system
Structured Output
Generate consistent, formatted meeting prep documents from AI synthesis
Graceful Degradation
Workflow delivers value even when some data sources are unavailable
Example meeting prep output
Here’s what a completed meeting prep document looks like:Customization ideas
Extend this workflow to match your sales process:Add CRM integration for deal history
Add CRM integration for deal history
Connect deeply with your CRM for complete context:
- Salesforce: Use Salesforce API to pull account, opportunity, and contact data
- HubSpot: Query deals, companies, and contact timeline
- Custom CRM: Build API integration with REST/GraphQL endpoints
- Complete interaction timeline (emails, calls, meetings)
- Support ticket history (if customer)
- Product usage analytics (if available)
- Sentiment analysis from past communications
- Win/loss analysis from similar deals
- Add “CRM Context Enrichment” agent before parallel research
- This agent queries CRM and structures data for other agents
- Include CRM data in all relevant research branches
Include social media research
Include social media research
Add competitor analysis branch
Add competitor analysis branch
Research competitive landscape:
- Add 6th parallel branch: Competitive Intelligence Agent
- Research competitors mentioned in CRM notes
- Find recent competitor news and product updates
- Compare positioning and messaging
- Identify competitive wins/losses in similar deals
- Competitor strengths and weaknesses
- Recent competitive news
- Differentiation talking points
- Battle cards and objection handling
Schedule automatic prep 24h before meetings
Schedule automatic prep 24h before meetings
Proactive meeting preparation:
- Integrate with calendar (Google Calendar, Outlook)
- Monitor calendar for new meetings with external companies
- Automatically trigger research workflow 24h before meeting
- Send meeting prep via email or Slack
- Include “Review before meeting” reminder
- Add calendar webhook trigger
- Parse meeting invitations for company names
- Schedule workflow execution for 24h before meeting time
- Add notification node to deliver prep document
Add industry-specific research
Add industry-specific research
Customize research by industry vertical:
- Create industry-specific research agents:
- Healthcare: HIPAA compliance, patient outcomes data
- Financial Services: Regulatory news, compliance requirements
- SaaS: Tech stack, integrations, scaling challenges
- Manufacturing: Supply chain, production capacity
- Add “Industry Classifier” agent at workflow start
- Route to industry-specific parallel branches
- Include industry-specific talking points
- Reference vertical-specific case studies
Build team collaboration features
Build team collaboration features
Share intelligence across your sales team:
- Save prep documents to shared repository
- Tag by company, industry, opportunity stage
- Enable search: “Show me past briefs on SaaS companies”
- Collaborative annotations: Team members add notes
- Feedback loop: Mark “useful” vs “not useful” insights
- Learn over time: Train synthesis agent on what works
Add real-time alert monitoring
Add real-time alert monitoring
Stay informed about customer developments:
- Create monitoring workflow (separate from prep workflow)
- Monitor news, social media, CRM activity for key accounts
- Send alerts when significant events occur:
- Funding announcements
- Executive changes
- Major partnerships
- Negative news
- Include suggested actions: “Reach out to congratulate on funding”
- Scheduled workflow (daily or real-time via webhooks)
- Monitor priority accounts
- Filter for high-relevance events
- Send Slack/email alerts to account owners
Best practices for sales intelligence
- Data Quality
- Privacy & Compliance
- Sales Enablement
- Continuous Improvement
Ensure accurate, relevant research:
- Validate company names before research (handle typos, abbreviations)
- Use high-quality sources (official websites, major news outlets)
- Include source citations for fact-checking
- Flag low-confidence information clearly
- Update research periodically (don’t rely on stale data)
- Handle data gaps gracefully (don’t hallucinate information)
Next steps
Now that you’ve built a sales intelligence assistant, explore related cookbooks:- RFP Proposal Analysis - Apply parallel research to proposal evaluation
- HR Policy Assistant - See another chat-triggered assistant pattern
- IT Support Triage - Another customer-facing automation workflow
Want to add advanced features like sentiment analysis, competitive battle cards, or predictive deal scoring? Contact our solutions team for guidance.