Grigoriy Dobryakov Courses Access

Courses Authors

AI Pilots to Production

Anatomy of Failure: Why 88% of AI Pilots Never Reach Production

○ The Illusion of Readiness: 1% vs. 99%
○ The 95% Trap: No Impact on P&L
○ Technical Debt by Default

The Business Contract and Baseline Lock-In

○ Forming the Contract Before Kickoff
○ Decomposing Success Metrics
○ The Budget Gateway

FinOps, Infrastructure, and AI Unit Economics

○ Scale Shock and TCO (Total Cost of Ownership)
○ Infrastructure SLAs and Cold Starts
○ Architectural Cost Optimization

Architectural Wrappers and the AI SDLC

○ Vendor Lock-in and LLM Gateways
○ Multi-Agent Chaos and A2A Protocols
○ Spec-Driven Development in AI
○ Continuous Evaluation
○ Deterministic Paths and Fallbacks

Data Engineering and Overcoming Entropy

○ The Death of Naive RAG and the Shift to GraphRAG
○ Training-Serving Skew
○ Integrating with Enterprise Legacy
○ Freshness and Conflict Resolution

Security-by-Design, Legal, and Compliance

○ Audit, Isolation, and Jailbreak Protection
○ Role-Based Access Control (RBAC) in Vectors
○ The Black Box Problem (Traceability & Audit Trail)
○ Legal Freeze, IP, and Regulation (EU AI Act / GDPR)

Strategy, Governance, and Ownership

○ The “SaaS Wipeout” Threat (Build vs. Buy)
○ Shadow AI and Decentralized Budgets
○ Committees vs. P&L (Passing the Baton)
○ The Responsibility Matrix (Governance Roles)

Business Process Reengineering (BPR) and AI Product UX

○ Chatbot Fatigue and UX Paradigms
○ Controls before Autonomy
○ Eliminating Manual Handoffs
○ Failure Protocols and Bounded Autonomy

Change Management and Team Transformation

○ Trust Penalty
○ Cognitive Overload and Present Bias
○ Designing Usage Guidelines
○ Workforce Readiness and Adoption Metrics

Go-to-Production: Launch Plan and Release

○ Pre-Release Architecture Audit
○ Roll-out Strategy (Shadow Mode)
○ Scaling and Transferring Autonomy

Data Engineering and Overcoming Entropy

Author

Grigoriy Dobryakov · Head of AI Engineering
Builds AI systems that run how engineering and delivery actually operate — hands-on engineer, not just a manager.

Sign in with Google to continue this module

Topics

The Death of Naive RAG and the Shift to GraphRAG
Training-Serving Skew
Integrating with Enterprise Legacy
Freshness and Conflict Resolution