Agent-led dev use case

Use Claude Code Or Codex To Ship Salesforce Improvements Safely

AI agents can now build useful Salesforce changes quickly, but without guardrails they can also create messy code, risky deployments, or false confidence.

Claude Code Salesforce development training Switzerland
Swiss SME workflow
Workshop-ready
Human approval by design
Before AI

How it looks today

  • Admins know the business problem but cannot safely make code changes.
  • Developers wait for detailed specs.
  • Small changes compete with larger project work.
  • Testing and review are inconsistent.
What AI can do

The first useful pass

  • Turn a task description into a branch and first implementation.
  • Draft LWC, Apex, Flow documentation, and test cases.
  • Explain the change for admin and developer review.
  • Iterate UI details quickly from feedback.
Humans approve

What stays in human hands

  • Architecture choices.
  • Permissions and data access.
  • Deployment to shared sandboxes or production.
Tools involved

The stack we usually meet

  • Claude Code
  • Codex
  • Salesforce CLI
  • GitHub
  • VS Code
  • Sandboxes

Workshop exercise

Agent-led Salesforce development

Set up the development environment, give the agent a contained Salesforce task, review the diff, run checks, and open a pull request.

Expected outcome

A practical operating model for agent-assisted Salesforce work that business teams and developers can trust.

Bring this to a workshop