AI Agent & Automation Workflows

Design Invocable Apex actions that return deterministic, agent-friendly errors instead of surfacing raw exceptions to the LLM. NOT for generic Apex exception handling or Flow fault paths.

Use when deploying an Agentforce agent to a channel surface: Embedded Service web chat, Slack, Experience Cloud, Agent REST API, or Salesforce Mobile. Triggers: 'deploy agent to website', 'add Agentforce to Slack', 'enable agent for mobile app', 'configure embedded service for agent', 'agent API for custom integration', 'multi-channel agent deployment'. NOT for agent logic design — use agentforce/agent-topic-design or agentforce/agent-actions for topics, instructions, and action contracts.

Use when writing or auditing the conversational copy layer of a Salesforce bot or Agentforce agent: utterance authoring strategy, fallback message copy, escalation-criteria phrasing, and persona-consistent dialog scripting across channels. NOT for platform configuration (bot builder setup, topic metadata wiring, action mapping, or Agentforce deployment) — see architect/einstein-bot-architecture and agentforce/agentforce-persona-design for those concerns.

Observability for Agentforce: adoption, deflection, latency, cost, quality. NOT for agent evaluation/testing (see agentforce-eval-harness) or raw platform-event monitoring.

Control LLM spend and Apex governor exposure for high-traffic Agentforce agents via per-user token budgets and graceful fallback. NOT for API rate-limiting of REST endpoints.

Pre-production security checklist for Agentforce deployments: permission scope, data exposure, authentication, logging. NOT for general Salesforce security review (see security-health-check).

Use when testing, evaluating, or building regression suites for Agentforce agents: conversation testing in Agent Builder, topic coverage and utterance testing, Testing API and AiEvaluationDefinition metadata, evaluation metrics (containment rate, escalation rate, CSAT, topic activation accuracy), and post-deploy analytics via Enhanced Event Logs. Triggers: 'how do I test my Agentforce agent', 'agent routes to wrong topic', 'write utterance tests', 'regression test after topic change', 'measure agent quality', 'agent containment rate'. NOT for agent creation, topic design, or action contract design — use agentforce/agentforce-agent-creation, agentforce/agent-topic-design, or agentforce/agent-actions respectively.

Use when designing or reviewing Agentforce topic structure, including topic boundaries, instruction quality, handoff rules, out-of-scope behavior, and topic-selector strategy. Triggers: 'agent topics', 'topic design', 'topic selector', 'agent scope boundary', 'handoff conditions'. NOT for action contract design or prompt-template wording when the main problem is not topic scoping.

Use when creating, configuring, auditing, or troubleshooting an Agentforce agent end-to-end: agent definition, agent user setup, channel assignment, system instructions, activation, and lifecycle management. Triggers: 'create agentforce agent', 'agent not appearing to users', 'how to activate agent', 'agent channel setup', 'agent lifecycle', 'deploy agent to production'. NOT for topic design or action contract design — use agentforce/agent-topic-design and agentforce/agent-actions respectively.

Use when designing how an Agentforce agent transfers the conversation to a human agent (Omni-Channel), to another bot/agent, or to an alternate workflow — including context package, deflection, escalation triggers, and user messaging. Triggers: 'agent to human handoff', 'agentforce escalate to omni channel', 'agent to agent handoff', 'transfer conversation with context', 'agent deflection fallback'. NOT for topic selector design (see agent-topic-design).

Use when Agentforce run costs are climbing, you need to forecast scale, or you want to reduce tokens per conversation without hurting quality. Covers topic design impact on cost, prompt/template reuse, grounding size discipline, caching, and model-tier selection. Triggers: 'agentforce cost', 'tokens per conversation too high', 'reduce agentforce runs spend', 'forecast agentforce scale cost', 'einstein trust layer tokens'. NOT for general LLM pricing strategy outside Salesforce.

Use when defining or refining the tone, voice, and behavioral personality of an Agentforce agent: system instruction encoding, brand voice alignment, adaptive response formats, multi-persona strategies. NOT for agent topic design (use agent-topic-design) or testing methodology (use agent-testing-and-evaluation).

Redact PII before it reaches Agentforce prompts, models, and logs. Trigger keywords: agentforce pii, pii redaction, data masking llm, einstein trust layer, prompt pii filter, audit pii leakage. Does NOT cover: Shield Platform Encryption at-rest (separate skill), GDPR data subject requests, or classic field-level security policy.

Version Prompt Templates and agent topic prompts: source-control shape, change review, model-version pinning, A/B, and rollback. Trigger keywords: prompt template versioning, prompt changelog, prompt rollback, A/B prompt test, agentforce prompt release. Does NOT cover: prompt engineering tips, general LLM fine-tuning, or Classify / Einstein Generate studio UI walkthroughs.

Step-by-step setup and configuration of Einstein for Sales AI features: Opportunity Scoring, Pipeline Inspection AI insights, Einstein email insights and composition, and Forecasting AI. Covers prerequisites, license checks, feature sequencing, and data readiness validation. NOT for core Agentforce agent creation, agent topic design, Einstein Trust Layer configuration, or Einstein Activity Capture troubleshooting.

Design Agentforce testing: topic coverage, action unit tests, deterministic golden sets, adversarial prompts, and regression harness. Trigger keywords: agentforce testing, agent eval, agent regression suite, prompt golden set, action unit test agentforce. Does NOT cover: generic LLM evaluation academia, human-labeled RLHF pipelines, or Einstein Classify accuracy.

System architecture for Salesforce AI governance: MLOps pipeline design, AI Audit Trail architecture, Einstein Trust Layer security design, Policy-as-Code engine, and regulatory compliance design for EU AI Act and similar frameworks. NOT for general Salesforce security architecture, AI ethics policy documentation, or individual Agentforce agent configuration.

Use when designing or auditing a Salesforce data architecture to support AI features — Einstein, Agentforce, Data Cloud, or custom ML models. Covers field-level data quality requirements, structured vs unstructured data placement, embedding-ready text field design, knowledge article structure for RAG grounding, Data Cloud harmonized data model, Einstein Feature Store patterns, grounding data for Agentforce actions, AI-ready field naming conventions, and data freshness requirements for model retraining. Trigger phrases: design data model for AI, make Salesforce data AI-ready, data quality for Einstein, grounding data for Agentforce, Data Cloud for AI, RAG data architecture Salesforce. NOT for Data Cloud initial setup and licensing, not for LWC or Flow implementation, not for general data migration, not for Einstein Analytics dashboard design.

Use when designing the architecture for AML/KYC compliance workflows on Salesforce — covering verification orchestration, third-party screening integration patterns, risk scoring design, and regulatory data flow. Triggers: AML architecture, KYC verification workflow design, sanctions screening integration, customer risk rating design, onboarding compliance workflow, watchlist screening Salesforce, FinCEN / FATF obligation data flow, BSA compliance architecture. NOT for implementing KYC data collection forms or FSC Action Plans (use admin/fsc-action-plans). NOT for configuring FSC Identity Verification for contact-center caller authentication. NOT for production screening vendor ISV configuration.

Use when designing or troubleshooting CRM Analytics dashboards — chart types, bindings, faceting, dashboard interactions, mobile layout, and filters. NOT for standard Salesforce reports and dashboards.

Use when creating or managing CRM Analytics datasets: configuring dataflows, scheduling refreshes, selecting fields, handling date types, managing row-count growth, or troubleshooting stale or broken datasets. Triggers: 'dataset creation', 'dataflow schedule', 'dataflow quota', 'date field not filterable', 'dataset row limit', 'CRM Analytics data refresh', 'field type mismatch in dataset'. NOT for standard Salesforce report types or choosing between analytics tools.

Strategy layer for resilient Apex HTTP callouts: bounded retry with backoff, queueable async retry chains, circuit-breaker via Platform Cache, idempotency keys, dead-letter pattern. NOT for callout authentication — see apex-named-credentials-patterns. NOT for transaction-boundary rules — see callout-and-dml-transaction-boundaries. NOT a re-write of the HttpClient template — this is the policy and orchestration layer that calls it.

Use when designing, reviewing, or debugging Apex code that relies on List, Set, or Map collections in triggers, batch classes, or service layers — especially for bulkification, heap management, and safe null handling. Trigger keywords: 'Map ', 'containsKey', 'retainAll', 'putAll', 'Set intersection', 'heap limit', 'collection in loop', 'unbounded accumulation'. NOT for SOQL query optimization (use soql-fundamentals), NOT for async job design (use apex-queueable-patterns or batch-apex-patterns), NOT for platform cache strategies (use platform-cache).

Use when diagnosing or preventing Apex CPU time and heap size problems, including nested-loop refactors, JSON memory pressure, string work, and Limits.getCpuTime() checkpoints. Triggers: 'CPU time limit exceeded', 'heap size too large', 'string concatenation', 'regex in loop', 'Limits.getCpuTime'. NOT for generic SOQL/DML governor-limit basics without a CPU or heap bottleneck.