The Salesforce dev world is unusual: governor limits, async patterns, metadata coupling, declarative-first culture. Most generic AI coding advice doesn't translate. This playbook is the opposite, it's specifically about what works for Salesforce developers, written by a team that's shipped AI-generated code into real production orgs.
What this playbook covers
We walk through five concrete workflows:
- Greenfield Apex, generating triggers, handlers, and service classes that respect bulk-safe patterns
- LWC scaffolding, components that compose with
lightning-record-edit-formandgetRecordcorrectly - Reading inherited code, using AI to understand the Apex you didn't write
- Test class generation, what AI does well, and the false-coverage trap to avoid
- Code review with AI, pre-PR review patterns that catch governor-limit and SOQL-in-loop issues
The core principle
AI is a force multiplier for what you already know. It's not a replacement for understanding governor limits, the order of execution, or the metadata API. The developers getting the most leverage are the senior ones, because they can spot when the AI is wrong fast enough that the speed gain still nets out positive.
This playbook assumes you already know Apex. It teaches you how to use AI to go faster without losing the code review discipline that keeps your org healthy.