$ Dispatch · AI Notes
How AI was used to build this piece, then how AI is used inside it.
## BUILD_PROCESS
Dispatch was built with Claude Code as a pairing partner throughout — scaffolding the Salesforce DX project structure, drafting Apex classes and Lightning Web Components against this site’s own proposal-then-drift-audit process (see Sreditor for that discipline in more depth), and iterating on the concurrency-safe claiming logic until the row-lock behavior was actually verified, not just assumed correct.
## AI_MATCHING_MECHANIC
JobMatchingServicequeries Contacts whose Service Region matches the Job’s Region, whose Certifications include the Job’s Job Type, and whose Availability Status is Available, sorts the shortlist by Rating, then calls Anthropic — via the Anthropic_APINamed Credential, never a hardcoded key — for a short natural-language reason per candidate. If no Contact is eligible, the service returns an empty result without ever calling Anthropic.
The contrast with Credential Flow:this callout originates from Apex inside Salesforce; Credential Flow’s originates from Python outside it. Same “explain the why” pattern, same underlying AI provider, opposite direction of integration — two genuinely different ways of proving an AI feature against Salesforce data.