AI Assisted Software Delivery Training Course

Module 1: Context, Scope, and Delivery Challenges

• Distinguishing between autocomplete features and autonomous multi-step execution
• Addressing common misconceptions regarding AI in software delivery
• Why superior prompting alone is insufficient for effective delivery
• Assessing participants' current tooling, pain points, and objectives
• Selecting the optimal AI operating model for engineering teams

Module 2: Specification Ingestion and Structured Decomposition

• Creating a structural inventory of stakeholder documentation
• Techniques for extracting precise requirements
• Chunking strategies: structural, semantic, and sliding-window approaches
• Maintaining dependencies and cross-references within content
• Processing tables, diagrams, flowcharts, and mixed media inputs
• Effective management of context windows

Module 3: Defining Human Judgment Boundaries

• Identifying areas where human decision-making remains essential
• Recognizing hallucinated dependencies in AI outputs
• Detecting fabricated constraints and logical inversions
• Mitigating risks from unsafe but seemingly helpful defaults
• Implementing validation frameworks for traceability, consistency, and completeness

Module 4: From Requirements to Code with Agentic Tools

• The architecture-first delivery model
• Mapping components and defining service boundaries
• Using API contracts as delivery anchors
• Establishing persistent rules and constraints within AI tools
• Linking specific task instructions to original requirements
• Comparing minimal prompting versus constrained prompting strategies
• Contract-first generation for both backend and frontend components

Module 5: The Agentic Iteration Loop

• The self-correction spiral concept
• Managing controlled iterative delivery cycles
• Reviewing diffs and code modifications
• Detecting scope creep and unauthorized changes
• Handling limited context memory effectively
• Leveraging iteration history for continuous improvement

Module 6: Enforcing Code Quality

• Applying prompt constraints to handle edge cases
• Treating rules documents as living governance artifacts
• Implementing automated gates using linting and static analysis
• Conducting security scans on AI-generated code
• Performing dependency and architecture conformance checks
• Protocols for human review of AI outputs

Module 7: Feedback Loops and Continuous Improvement

• Incorporating structured failures back into AI workflows
• Defining bounded iterations and clear stop criteria
• Logging cycles and analyzing outcomes
• Refining rules documents over time
• Building reusable engineering intelligence

Module 8: Security Anti-Patterns in AI Delivery

• Identifying common security risks in generated code
• Applying technology-specific security rule appendices
• Implementing pre-commit security scanning
• Integrating Secure SDLC controls for AI-assisted development
• Ensuring human accountability in secure delivery processes

Module 9: Testing Anchored to Specifications

• Deriving test specifications directly from requirements
• Designing tests using domain language
• Safely generating test implementations
• Concepts in mutation testing
• Validating specification coverage
• Reviewing assertion strength
• Utilizing diagnostic questioning models

Module 10: System Maintenance

• Managing living artifacts: contracts, maps, rules, and test specs
• Evolving constraints as the system matures
• AI governance for long-term maintainability
• Preventing technical debt through AI controls
• Establishing an operating model for sustainable AI engineering teams