👥 Participants
Yoni Szpiro
Co-founder, Generations RPO
New ABChat user
David Caggiari-Pallett
Co-founder, Generations RPO
Existing ABChat power user
Caroline Hayat
Consultant
Interested in media monitoring use case
Giobi Fasoli
ABChat Creator
Demo presenter
⚙️ Technical Setup Completed
Yoni's Workspace Live
- Workspace created & configured
- User linked: [email protected]
- Nginx routing configured
- VS Code dark theme applied
- SAGE agent configured (French primary)
- Script 'start' operational
Issues Resolved Fixed
- Database user ID mismatch (ID 8 vs 9)
- David's start script updated
- BOOTSTRAP translated to French
- Workspace visibility in dashboard
⚠️ Important: Yoni needs to setup his own Claude Pro/Max account to avoid token limits on Giobi's account.
🔍 Claude Cowork vs ABChat Brain
Key question from Caroline: "What's the real difference?"
Claude Cowork
- Access to ONE local folder
- Simplified UI for non-developers
- Limited agency
- macOS app only
- Pre-configured connectors
- No bash/SSH commands
- No automation/cron
- Session-based memory
ABChat Brain
- Full server filesystem access
- Bash/SSH/git commands
- Structured local database
- Automation & cron jobs
- Custom Python wrappers
- Persistent memory (diary/ + MEMORY.md)
- Fully configurable personality
- Multi-tenant architecture
- Unlimited API integrations
Analogy: Cowork = Assistant with access to one office. Brain = Assistant + Full office + Secretariat + Automation infrastructure.
📰 Caroline's Use Case: Media Monitoring
Requirements
- Scrape 5,000+ links from European media
- Extract data into Excel (20 columns: author, date, theme, etc.)
- Monitor mentions of EU agencies (e.g., EIGE)
- Currently done manually for 5 years
Live Demo Results
- Real-time scraping demonstrated
- 59 initial results found (3 months)
- Deep scraping possible with more time
- Can bypass some paywalls (white hat)
- ~90% coverage estimated
Caroline's Feedback
- Very interested but skeptical
- Previous bad experiences with developers
- Needs exhaustive results (100%)
- Wants dedicated follow-up call
- Requests test with real dataset
🔒 GDPR & AI Act Compliance
Critical Discussion: Caroline raised important regulatory concerns that must be addressed for RPO/HR use cases.
Concerns Raised
- GDPR critical for RPO/HR sector
- AI Act regulations vary by use case
- Anonymization alone insufficient
- Data crossreferencing can identify people
- Trade union resistance to AI
- Shared liability issues
Recommendations
- Don't trust generic "respect GDPR" commands
- Due diligence per use case
- Specific testing required
- Legal specialist consultation
- Human oversight mandatory
- MSA with specific IP/compliance clauses
Team Response
Agreed on:
- Agents need detailed configuration, not generic rules
- Legal oversight essential
- "White hat" approach only
- Human-in-the-loop always present
- Caroline's expertise valuable for compliance framework
💼 Business Model Discussion
Generations RPO Strategy
Core Concept: Sell customized agents as a service (not the Brain itself)
- Client audit & use case identification
- Create specific agents (BizDev, Legal, Scraper, etc.)
- Industry/company customization
- MSA with IP, compliance, liability clauses
- Digital outsourcing positioning
IP Strategy: Core framework reusable. Client-specific customization is their IP. No reselling specific agents to competitors.
🎯 Action Items
- Yoni: Setup Claude Pro/Max account + complete BOOTSTRAP onboarding
- Giobi: Test Claude Cowork thoroughly for accurate comparison
- Giobi: Create systemd service for Yoni's opencode
- Giobi: Prepare UI-friendly Brain version
- Giobi: Draft MSA template with legal language
- David: Continue exploring RPO use cases, provide feedback
- Caroline: Schedule dedicated call for media scraping demo
- Caroline: Provide test dataset (5,000 links) for exhaustiveness testing
- Team: Define compliance checklist for agents
- Team: Explore Caroline as compliance advisor role
💡 Key Insights
Product Maturity
- Brain becoming production-ready
- Portability is strong selling point
- No vendor lock-in (can migrate to any LLM)
- UI improvements needed for non-technical users
Market Positioning
- "Agent as a service" model viable
- Focus on "Eurocentric, compliant" positioning
- Compliance framework essential for credibility
- Human-led service differentiator
User Adoption
- David: Power user, high enthusiasm
- Yoni: Cautious but interested (appropriate for HR)
- Caroline: Skeptical but willing to test
- Onboarding process needs refinement
Risks & Challenges
- Regulatory compliance critical
- Cannot do "cowboy" implementations
- Token costs significant for heavy use
- Technical barrier for non-developers