PACE Pattern Overview

Complete Guide to Pattern for Agentic Conversational Experience

---

What is PACE?

PACE (Pattern for Agentic Conversational Experience) is a UX architecture pattern that replaces traditional browse-and-click interfaces with guided conversations.

Instead of presenting users with navigation, grids, and filters, a PACE implementation uses an AI guide to lead users through dialogue.

---

The Core Concept

Traditional UX: User → Browse → Filter → Search → Decide
PACE UX: User → Guide → Conversation → Recommendation → Decide

Key difference: The guide actively pursues user intent instead of passively presenting options.

---

The PACE Acronym (Two Meanings)

Layer 1: What It Is

P.A.C.E. = Pattern for Agentic Conversational Experience

A framework for building guide-first interfaces.

Layer 2: How It Behaves

P.A.C.E. = Proactive, Adaptive, Contextual, Efficient

The four behavioral principles every PACE implementation must embody.

Learn more →

---

The Four Components

Every PACE implementation has these components:

Component	Purpose	Example
Product	AI-guided catalog	"Here are 3 MCP servers that fit your needs"
About	Context and trust	Origin story, philosophy, credibility
Chat	Conversational interface	Primary dialogue with the guide
Executive Summary	Real-time insights	"You've discussed SQL MCP and Schema Explorer"

Explore Components →

---

The Four Principles

Every PACE guide should be:

1. Proactive

Don't wait. Initiate.

❌ Empty search box
✅ "Welcome to the pond. What are you fishing for?"

2. Adaptive

Match the user's level.

Beginner: "Think of MCP servers like plugins for Claude"
Expert: "Implements JSON-RPC 2.0 over stdio transport"

3. Contextual

Remember and reference.

"You mentioned earlier you're using PostgreSQL..."
"Circling back to your question about pricing..."

4. Efficient

Be concise. Be actionable.

❌ "Well, there are many options available..."
✅ "Here are 3 options: [list with CTAs]"

Deep dive into Principles →

---

Problem & Solution

The Problem: Traditional Product Discovery

Issue	Impact
Cognitive overload	Too many options paralyze decisions
Hidden gems	Best products buried in catalogs
Context loss	Each page view loses conversation thread
Generic UX	Same experience for beginners and experts
Documentation required	Users must read before understanding

The Solution: Guide-First UX

Traditional	PACE
Browse 50 products	"What are you looking for?"
Search + filter	Natural language query
Read docs first	Guide explains as needed
One-size-fits-all	Adapts to expertise
Static grid	Dynamic conversation

---

Architecture Diagram

---

How PACE Works

1. Guide Initiates

Guide: "Welcome to the pond. What are you fishing for?"

Why: Users don't know where to start. The guide breaks the ice.

2. User Expresses Intent

User: "I need something to help with database queries"

Why: Natural language is easier than navigating menus.

3. Guide Detects Expertise

Analysis:
- No technical jargon → likely beginner
- Mentions "database" → has some context
- Adapt: friendly + practical

Why: One size doesn't fit all. Beginners need different guidance than experts.

4. Guide Recommends

Guide: "For database work, I'd recommend:

1. **SQL MCP** — Query databases with natural language
2. **Schema Explorer** — Visual database browsing
3. **Query Builder** — Interactive SQL generation

Which sounds most useful?"

Why: Curated options reduce cognitive load. Questions invite engagement.

5. Conversation Continues

User: "Tell me more about SQL MCP"

Guide: "SQL MCP connects Claude to your database.
You can ask questions like 'Show me users who signed up last week'
and it generates the SQL query for you.

[Try it] [View docs] [See examples]"

Why: Conversation builds understanding progressively.

6. Executive Summary Tracks Progress

Products discussed: SQL MCP, Schema Explorer
User expertise: Beginner
Suggested next step: Try SQL MCP

Why: Meta-awareness helps users see progress and next steps.

---

When to Use PACE

PACE is ideal for:

✅ Product catalogs — Help users find the right product ✅ Documentation sites — Guide users to relevant docs ✅ SaaS onboarding — Personalized setup guidance ✅ E-commerce — Curated shopping experiences ✅ Service selection — Match users with right services

PACE may not fit:

❌ Pure content consumption — Blogs, news sites ❌ Known-item search — User knows exactly what they want ❌ High-frequency tasks — Repeated daily actions (email, chat) ❌ Exploration-focused — User wants to browse randomly

---

Implementation Approaches

Option 1: PACE.js Framework

import { PACE } from '@semanticintent/pace-pattern'

const pace = new PACE({
  container: '#app',
  products: './products.json',
  aiAdapter: 'claude'
})

pace.mount()

Pros: 15KB, zero dependencies, framework-agnostic

View Framework Docs →

Option 2: Custom Implementation

Build your own using:

• React + LangChain
• Vue + OpenAI SDK
• Django + Claude API
• Your stack + Any LLM

Pros: Full control, integrate with existing architecture

See Integration Guide →

Option 3: Hybrid

Use traditional UX for most of the site, PACE for specific sections:

Homepage: Traditional grid
Product detail: PACE guide for recommendations
Checkout: Traditional flow
Support: PACE guide for help

---

Measuring Success

Metric	Traditional UX	PACE UX
Time to decision	High (browsing all options)	Lower (guided recommendations)
Bounce rate	High (overwhelmed users leave)	Lower (engaged in conversation)
Conversion rate	Variable	Higher (personalized guidance)
User satisfaction	Depends on findability	Depends on guide quality
Cognitive load	High (many choices)	Lower (curated options)

---

Key Takeaways

1. PACE is a pattern, not a product — Implement it your way
2. Four components are required — Product, About, Chat, Executive Summary
3. Four principles are non-negotiable — Proactive, Adaptive, Contextual, Efficient
4. Guide quality determines success — Better AI = better experience
5. Not for every use case — Choose when guidance adds value

---

Quick Reference

┌─────────────────────────────────────────────┐
│              PACE Pattern                    │
├─────────────────────────────────────────────┤
│                                             │
│  WHAT: Pattern for Agentic Conversational   │
│        Experience                           │
│                                             │
│  HOW: Proactive, Adaptive, Contextual,      │
│       Efficient                             │
│                                             │
│  COMPONENTS: Product, About, Chat,          │
│              Executive Summary              │
│                                             │
│  MANTRA: "Ask, don't browse"                │
│          "Conversation over catalog"        │
│          "Guide-first, not grid-first"      │
│                                             │
└─────────────────────────────────────────────┘

---

Next Steps

Learn the Pattern

• Study the Components — Product, About, Chat, Executive Summary
• Master the Principles — Proactive, Adaptive, Contextual, Efficient
• Explore the Matrix — 3D semantic structure
• Read the Origin Story — How PACE came to be

Build with PACE

• PACE.js Framework — 15KB JavaScript implementation
• Quick Start — Build your first PACE app
• MillPond Example — See a complete implementation
• Full Tutorial — Step-by-step guide

---

Don't make users hunt. Let the guide fish for them. 🐦