Admin, Development, and Automation

Section 7: Use in Salesforce Admin Work

7.1 High-value admin use cases

• field creation planning
• validation rule drafting
• formula generation
• Flow design and naming
• Flow debugging
• page layout and Dynamic Forms recommendations
• report/dashboard definitions
• permission set design support
• email template drafting
• assignment and escalation rule logic
• knowledge setup
• case configuration
• sandbox setup checklists
• release note summarization

7.2 Examples

#### Validation rule

Create a Salesforce validation rule for Opportunity that prevents stage moving to Closed Won when Primary_Contact__c is blank and Implementation_Date__c is in the past. Explain the logic in plain English and include test scenarios.

#### Formula support

Create a Salesforce formula field that classifies a case as SLA Breached if Status is not Closed and NOW() is greater than Milestone_Target__c. Also provide null handling guidance.

#### Flow debugging

Review this screen flow outcome summary and explain likely failure points:
- user sees generic fault message after submit
- records created sometimes duplicate
- one path skips approval task creation

Recommend debug steps, logging strategy, and design fixes.

7.3 Admin best practices

• Always ask AI for both formula/Flow logic and plain-English explanation.
• Request test scenarios with every admin artifact.
• Ask for naming conventions, descriptions, and maintenance notes.
• Review performance impact before implementing complex Flow loops.

7.4 Do and don't

Section 8: Use in Apex Development

8.1 Where AI is most valuable

AI is highly effective in Apex when used for:

• class scaffolding
• trigger/handler structure
• service layer and selector patterns
• queueables, batches, schedulables
• invocable methods
• REST resources
• callout code
• platform event subscribers
• wrapper classes
• test data setup and test classes
• refactoring and readability improvements
• code review and defect isolation

8.2 Target patterns

• trigger framework
• domain/service separation
• selector pattern for SOQL centralization
• utility classes
• custom metadata-driven logic
• error normalization
• retry strategies

8.3 Example scenarios

#### Contact lookup by email/phone

Use AI to generate:

• normalized search utility
• bulk-safe lookup service
• test coverage for duplicate hits and no-hit scenarios

Prompt

Create Apex using a service layer and selector pattern to find contacts by normalized email or phone.
Requirements:
- bulk-safe
- with sharing
- null-safe
- return wrapper with match confidence and source field
- include tests
- avoid SOQL in loops

#### Loan eligibility service

Use AI to:

• create a service class that evaluates applicant data
• externalize thresholds into custom metadata
• design a response wrapper for UI and integration reuse

#### External API callout

Use AI to:

• create named credential-based callout service
• add timeout handling and response parsing
• generate HttpCalloutMock tests

#### Batch data cleanup

Use AI to:

• generate a batch class to normalize duplicate data or archive stale records
• design retry/resume pattern where appropriate

#### Case assignment engine

Use AI to:

• compare Flow, custom metadata, and Apex routing options
• implement a metadata-driven assignment service

#### Reusable utility framework

Use AI to:

• refactor repeated date formatting, error handling, or mapping logic into a utility layer

8.4 Apex review checklist

• Is the code bulkified?
• Is sharing behavior correct?
• Are limits considered?
• Are SOQL/DML in loops avoided?
• Are nulls handled explicitly?
• Are callouts mocked in tests?
• Is error handling user-safe and support-friendly?
• Is custom metadata used where hard-coded business rules should not exist?

8.5 Human review required

Never deploy AI-generated Apex without:

• peer review
• static analysis
• governor limit review
• positive and negative test execution
• security review if handling sensitive data

8.6 Anti-patterns

• gigantic all-in-one triggers
• hidden side effects in helper methods
• hard-coded IDs and endpoints
• tests that only assert non-null
• logic that ignores mixed bulk operations

Section 9: Use in LWC Development

9.1 LWC use cases

• component scaffolding
• HTML/JS/CSS generation
• wire adapter usage
• imperative Apex calls
• event communication
• reusable components
• datatable patterns
• wizard flows
• modals
• file upload
• responsive layouts
• accessibility improvements
• Jest tests
• debugging rendering issues

9.2 Example use cases

#### Intake form

AI can draft:

• form sections
• client-side validation
• Apex submission contract
• UX copy

#### Contact creation wizard

AI can generate:

• multi-step navigation
• validation summary banner
• save/resume behavior
• toast/error handling

#### Dashboard widgets

AI can produce:

• KPI cards
• chart data transformation
• refresh logic
• empty state UX

#### Document upload screen

AI can support:

• file validation
• progress display
• upload error messaging

#### Custom portal component

AI can help with:

• Experience Cloud constraints
• external user-friendly messaging
• responsive design for smaller form factors

9.3 LWC prompt example

Create a Lightning Web Component for a 3-step customer intake wizard.
Requirements:
- mobile responsive
- accessible labels and keyboard navigation
- validate email, phone, and required documents
- use imperative Apex save on final step
- show inline errors and a summary banner
- include Jest tests outline

9.4 Best practices

• Ask for accessibility considerations every time.
• Ask for performance considerations when using datatables or repeated renders.
• Request both component code and explanation of state transitions.
• Validate security, field access, and data exposure assumptions.

Section 10: Use in Flow and Automation

10.1 Flow-related uses

• decision logic drafting
• naming convention enforcement
• screen copy/help text
• auto-launched flow planning
• subflow decomposition
• error path design
• failed flow troubleshooting
• Flow documentation
• process-to-Flow conversion
• Flow versus Apex decisions

10.2 Prompt example

Design a record-triggered flow for Case that:
- runs on create and update
- routes high-priority cases to an escalation queue
- creates a follow-up task when customer SLA risk is high
- avoids recursion
- includes fault paths
- uses maintainable naming conventions

Provide the recommended elements, decisions, entry criteria, and documentation notes.

10.3 Flow governance guidance

• prefer before-save for simple field updates
• use subflows for repeated business logic
• avoid overly complex loops in large-volume transactions
• document every fault path
• define ownership for every major automation

10.4 Flow vs Apex decision table