sf-ai-agentforce-conversationdesign（SF AI AgentForce 对话设计）

SF-AI-Agentforce-ConversationDesign Skill

"Users don't fail conversations — conversations fail users."

Conversation design is the discipline of crafting agent interactions that feel natural, resolve issues efficiently, and gracefully handle the unexpected. This skill brings structured conversation design methodology to Salesforce Agentforce, combining industry frameworks (Google, IBM, PatternFly) with Salesforce-specific implementation patterns.

⚡ Quick Start

New agent? Start here:

1. Design your persona → Persona Design Guide
2. Architect your topics → Topic Architecture Guide
3. Write instructions → Instruction Writing Guide
4. Score your design → Quality Scorecard

Existing agent needs improvement? Start here:

1. Run the Quality Scorecard assessment
2. Review Anti-Patterns for quick wins
3. Build an Improvement Plan

📚 Document Map

Tier 1 — Start Here

Tier 1.5 — Reference Guides (Extracted)

Tier 2 — Design Guides

Tier 3 — Reference Resources

Tier 4 — Templates & Examples

🏆 Scoring System (120 Points)

See references/scoring-rubric.md for the full per-criterion breakdown of all 8 categories.

Category Summary

Grade Scale

🎭 Persona Design

A persona defines your agent's personality, communication style, and behavioral constraints — the foundation for consistent, brand-aligned interactions.

Persona Components

1. Identity — Name, role, department, target audience
2. Tone Register — Casual, neutral, or formal (Agentforce setting)
3. Personality Traits — 3-5 traits that shape response style
4. Communication Style — Sentence length, vocabulary level, empathy patterns
5. Limitations — What the agent explicitly will not do
6. Messages — Welcome message and error/fallback message (<=800 chars each)

Salesforce Implementation

Agent Builder → Agent Settings → Instructions (Agent-Level)
Agent Builder → Agent Settings → Tone (Casual/Neutral/Formal)
Agent Builder → Channels → Welcome Message / Error Message

The persona lives primarily in agent-level instructions. Write them like training a new employee on Day 1 — focus on who they are and how they communicate, not on specific task procedures.

Deep Dive: Persona Design Guide | Template: Persona Document | Example: Service Agent Persona

🏗️ Topic Architecture

Topics group related actions under classification descriptions for routing. Use bottom-up design: list all actions → group by user intent → write classification descriptions → test for semantic distinctness.

Architecture Rules

Classification Descriptions

The single most important text in your agent design — they determine routing accuracy.

✅ GOOD: "This topic handles questions about existing order status, including
tracking information, estimated delivery dates, and order modification
requests. It does NOT handle new order placement or returns."

❌ BAD: "Order stuff"

Test: Can you read two descriptions and immediately tell which utterance belongs where?

Deep Dive: Topic Architecture Guide | Template: Topic Architecture | Example: Retail Topic Architecture

✍️ Instruction Writing

Instructions operate at three levels: Agent-level (persona, global rules, 200-500w), Topic-level (workflow logic, 100-300w), Action-level (when/how to invoke, 50-150w).

Core Principles

1. Guidance over determinism — Guide reasoning, don't hard-code every decision. "Prioritize empathy and escalate if unresolved within 2-3 exchanges" vs scripting every anger keyword.
2. Positive framing — "Always verify identity before accessing account details" not "Don't ever access without verifying."
3. Business principles, not decision trees — If your instruction contains if...then...else with thresholds, it belongs in Flow/Apex.
4. Knowledge over hard-coding — Use Knowledge actions (RAG) for policies, don't inline policy text in instructions.

Deep Dive: Instruction Writing Guide

🔄 Dialog Flow Patterns

Every conversation follows a six-phase lifecycle: Greeting → Classification → Gathering → Processing → Response → Close.

Key design rules:

• Progressive disclosure: Max 2-3 choices per turn. If more options exist, ask a qualifying question first.
• Context preservation: Reference prior turns, avoid re-asking for information already provided.
• Error recovery: Define clarification prompts, disambiguation paths, and graceful fallbacks for every phase.
• Explicit closing: Summarize what was accomplished, offer follow-up, farewell appropriate to tone.

📝 Utterance Design

Utterances are test cases for your topic architecture — they validate that classification descriptions route correctly.

Categories & Targets

Process: Start with real data (CRM cases, chat logs) → brainstorm synonyms → add edge cases → include adversarial → test in Testing Center → iterate on failures.

Template: Utterance Library

🚨 Escalation Design

Escalation is not failure — it's a safety net ensuring customers always reach resolution.

Trigger Catalog

When escalating, pass: conversation transcript, customer identity, issue summary, actions taken, and escalation reason. Agentforce provides a pre-built Escalation Topic that routes via Omni-Channel.

Deep Dive: Escalation Patterns | Template: Escalation Matrix

🛡️ Guardrails & Safety

Safety operates through four layers: Einstein Trust Layer (platform — toxicity detection, PII masking, automatic), Topic Classification (scope boundaries as first defense), Instructions (behavioral constraints — "never provide legal/medical advice"), Flow/Apex (deterministic business rule enforcement).

Critical Rule: Never rely on instructions alone for safety-critical decisions. Instructions are probabilistic. Financial limits, compliance checks, and approval gates MUST be in Flow or Apex.

Deep Dive: Guardrail Hierarchy

📊 Quality Assessment & Continuous Improvement

See references/quality-assessment.md for the full assessment process and 5-question quick health check.

See references/continuous-improvement.md for the Monitor → Analyze → Fix → Retest → Deploy iteration cycle, KPI targets, and utterance analysis process.

⚠️ Anti-Patterns

Deep Dive: Anti-Patterns — Full examples with before/after fixes.

🔗 Chain Integration

This skill is the first step in the Agentforce development chain:

📎 Credits & References

• Google Conversation Design Guidelines
• IBM Natural Conversation Framework
• Red Hat PatternFly AI Design System
• Salesforce Conversational AI Design Guide
• Salesforce Architect: Agentic Patterns & Taxonomy

See CREDITS.md for full attribution.