Applying AI-assisted development best practices

AI Assistant produces the best results when requests clearly describe the intended outcome, relevant constraints, and expected actions. Although AI Assistant can work with business language and ask clarifying questions, well-structured requests typically produce more accurate and useful responses. 

When you create requests, apply the following practices: 

• Define a clear objective so AI Assistant can focus on a specific outcome rather than inferring intent. 
• Provide relevant business and technical context, including requirements, constraints, and existing application behavior. 
• Establish boundaries by identifying what should remain unchanged and any implementation approaches to avoid. 
• Specify the expected action, such as analyzing, planning, implementing, reviewing, verifying, or generating tests. 
• Break large activities into smaller checkpoints so that you can validate progress and refine requirements before additional work begins. 

AI Assistant performs best when development activities follow a structured and iterative approach. Rather than attempting to implement a complete solution in a single interaction, use AI Assistant to progressively refine requirements, evaluate design options, and implement manageable units of work. 

Consider the following practices: 

• Use Plan mode to explore design alternatives and develop an implementation approach before making application changes. 
• Remain in Plan mode until requirements, assumptions, and design decisions are well understood. 
• Break larger initiatives into smaller user stories or implementation tasks that can be completed independently.  
• Build dependent Rules from the bottom up. When Rules depend on other Rules, implement the lowest-level Rule first. For example, if a Flow calls a Flow Action and the Flow Action references a View, create the View first, followed by the Flow Action, and then the Flow.  
• Ask AI Assistant to analyze existing Rules and application assets before directing it to create new artifacts. This approach can improve reuse and reduce duplication. 
• Review proposed designs, implementation plans, and generated artifacts before accepting them. AI-generated output can be incomplete, inaccurate, or inconsistent with application requirements. 

As an LSA, use AI Assistant to accelerate analysis and implementation activities, but continue to evaluate generated content against architectural principles, reuse opportunities, security requirements, application layering, and governance standards. 

Long-running conversations can gradually lose focus as objectives change and additional context accumulates. Effective session management helps maintain clarity and improves the quality of AI-assisted development activities. To keep sessions productive: 

• Focus each session on a single objective or a small group of related tasks. 
• Start a new session when work shifts to a different problem or design area. 
• Use recap points during longer sessions to summarize what you have completed, what remains in scope, and what outcome to address next. 
• Watch for signs of context drift, such as repeated responses, missing details, or unrelated recommendations. Starting a new session is often more effective than attempting to recover an increasingly complex conversation. 
• End sessions at logical stopping points, such as after completing a plan, implementing a user story, resolving a problem, or reviewing a design. 

For more information about best practices in AI- Assisted development, see Best practices for AI Assistant. 

Check your knowledge with the following interaction: