Cursor workflow optimization and EP task automation strategy

Date: February 12, 2026 Source: Granola Meeting ID: 5d7a9b0f-26e0-4241-9070-4fee20ceee33 URL: https://notes.granola.ai/t/5d7a9b0f-26e0-4241-9070-4fee20ceee33

Participants:

  • Uttam Kumaran (Brainforge)

Summary

AI Agent Development Framework

  • Creating “soul documents” for AI agents
    • Constitution defining behavior, functions, and core purpose
    • Similar to Claude’s system prompts but more comprehensive
    • Can be enriched with situational examples and interconnects
  • Building CSO, SL, and EP agent personas
    • Digital personas for bouncing ideas and task execution
    • Foundation for eventual pure agent automation
    • Current focus on tooling and NCP connections

Case Study Generation Process

  • Successfully created marine parts ecommerce case study
    • Used existing pool parts project as template
    • Generated aligned proposal content for IL team
    • Marketing team will use as template for AI-assisted drafting
  • Improvements made with pool parts repo access
    • Added actual data modeling logic and technical details
    • Enhanced shipping-focused content based on transcript history
    • Included source references for future case study development

EP Task Automation Strategy

  • Current EP responsibilities taking 60-90 minutes weekly
    • Slide deck updates, Linear ticket management, client status tracking
    • Target: reduce to 15 minutes per task through AI assistance
  • Three-phase approach identified:
    1. Get humans executing tasks efficiently with AI tools
    2. Build agent systems to handle routine tasks
    3. Full automation with human oversight for exceptions
  • Foundation requirements before automation:
    • Unified client hubs with all project resources
    • Connected data sources (Notion, Linear, Slack, Gantt charts)
    • Standardized processes and documentation

System Integration Challenges

  • Scattered client resources across multiple platforms
    • Instagant for project timelines
    • Linear for individual tickets
    • Notion for client info
    • No unified access point for comprehensive project view
  • MCP connection reliability issues
    • Fragile connections requiring monitoring and auto-repair
    • Need for health checks and error logging
    • OpenTelemetry recommended for production systems

Cursor Adoption and Training

  • Onboarding friction identified as major barrier
    • Requires hands-on teaching for effective adoption
    • People default to familiar tools when setup is difficult
    • Need to reduce options and force Cursor as primary tool
  • Usage patterns and best practices
    • Using Cursor as digital notebook with voice transcription
    • Auto model routing needed to optimize costs ($600/month current spend)
    • Session history recovery and management workflows

Production AI Review System

  • GitHub action for automated PR risk profiling
    • Labels: raw/half-baked/fully-baked, priority levels, effort estimates
    • Auto-generates descriptions and runs security/test checks
    • Enables Sam to focus on complex reviews, auto-approve simple changes
  • Scaling challenge: 30 PRs currently → 300 PRs projected
    • Need AI to handle 90% of routine reviews
    • Human reviewers focus on security, major changes, subject matter expertise

Next Steps and Priorities

  • Create comprehensive EP task inventory
    • Map all responsibilities with time estimates and difficulty ratings
    • Stack rank by impact and implementation feasibility
    • Focus on hour-to-15-minute reductions, not 15-to-5-minute optimizations
  • Establish 100% data freshness standards
    • 10 core questions every EP must answer instantly
    • Client hub accessibility test as pass/fail criteria
    • Tie performance bonuses to ticket freshness and documentation quality
  • Timeline: 2-3 weeks for EP foundation, then expand to CSO/SL roles