跳至正文
OpenAgentSkill
浏览已开放 · 目录记录日期 2026-11-21 · 自动更新和账户服务暂时关闭。
AI 与知识库

ai-native-product

Build AI-native products with agency-control tradeoffs, calibration loops, and eval strategies. Use when building AI agents, LLM features, or products where AI handles user tasks autonomously. Part of the Modern Product Operating Model collection.

YannickMaurice · 已有静态检查记录

安装与使用

运行命令前先阅读来源与依赖说明。复制命令不会自动安装或执行。

npx skills add yannickYamo/skills --skill ai-native-product

技能说明


name: ai-native-product description: Build AI-native products with agency-control tradeoffs, calibration loops, and eval strategies. Use when building AI agents, LLM features, or products where AI handles user tasks autonomously. Part of the Modern Product Operating Model collection. author: YannickMaurice version: 1.0.0 tags: product-management, ai, llm, agents, calibration, evals

AI-Native Product Development

"AI products aren't deterministic. They require continuous calibration, not just A/B tests."

This skill covers AI-Native Product Development — the overlay that modifies discovery, architecture, and delivery when AI is at the core. It addresses the unique challenges of building products where AI agents perform tasks autonomously.

Part of: Modern Product Operating Model — a collection of composable product skills.

Related skills: product-strategy, product-discovery, product-architecture, product-delivery, product-leadership


When to Use This Skill

Use this skill when:

  • Building AI agents that act on behalf of users
  • Adding LLM-powered features to existing products
  • Designing human-AI interaction patterns
  • Deciding how much autonomy to give AI
  • Setting up eval strategies and calibration loops
  • Managing the "agency-control tradeoff"

Not needed for: Traditional software products, ML models used only for backend optimization (no user-facing autonomy)


What Makes AI Products Different

Traditional Software vs. AI Products

DimensionTraditional SoftwareAI-Native Products
BehaviorDeterministicProbabilistic
TestingUnit tests, QAEvals, calibration
CorrectnessBinary (works or doesn't)Spectrum (good enough?)
User roleOperatorDelegator + Reviewer
Failure modeError messagesPlausible but wrong outputs
IterationShip → Measure → IterateShip → Observe → Calibrate
Trust buildingFeature completenessDemonstrated reliability

The Core Challenge

AI products must navigate a fundamental tension:

More autonomy = More value (fewer steps, faster outcomes)
More autonomy = More risk (errors affect real work)

This is the Agency-Control Tradeoff.


Framework: The CCCD Loop

Credit: Aishwarya Goel & Kiriti Gavini

AI products require a Continuous Calibration and Confidence Development (CCCD) loop:

┌─────────────────────────────────────────────────────────────────┐
│                        CCCD LOOP                                │
│                                                                 │
│    CALIBRATE → CONFIDENCE → CONTINUOUS DISCOVERY → CALIBRATE   │
│         ↓           ↓              ↓                 ↓         │
│     Eval and    Build user    Observe AI       Update evals    │
│     adjust AI    trust over   interactions     and models      │
│     behavior     time         at scale                         │
└─────────────────────────────────────────────────────────────────┘

CCCD Components:

ComponentPurposeActivities
CalibrateTune AI behavior to match user expectationsRun evals, adjust prompts/models, set guardrails
ConfidenceBuild appropriate user trustShow AI reasoning, enable verification, demonstrate reliability
Continuous DiscoveryObserve AI-user interactions at scaleLog interactions, identify failure patterns, surface edge cases
→ Back to CalibrateUpdate based on learningsImprove evals, retrain, adjust prompts

The Agency-Control Progression

Five Levels of AI Agency

LevelDescriptionAI DoesUser DoesExample
1. AssistAI suggests, user executesGenerates optionsChooses and actsAutocomplete, suggestions
2. RecommendAI ranks, user approvesAnalyzes and recommendsReviews and approves"AI recommends these 3 actions"
3. Execute with confirmationAI acts after approvalPrepares actionConfirms before execution"Send this email?" → Yes/No
4. Execute with notificationAI acts, notifies afterActs autonomouslyReviews outcomes"I scheduled the meeting and sent invites"
5. Fully autonomousAI acts without notificationHandles end-to-endSets goals, reviews exceptionsAI handles routine tasks silently

Progression Strategy

Start lower, earn higher:

Level 1 → Build trust → Level 2 → Demonstrate reliability → Level 3 → ...

Graduation Criteria:

From LevelTo LevelRequires
1 → 2Assist → RecommendUser accepts suggestions > 70%
2 → 3Recommend → Execute with confirmUser approves recommendations > 80%
3 → 4Execute+confirm → Execute+notifyUser confirms without edit > 90%
4 → 5Execute+notify → AutonomousUser overrides < 5%, high-stakes scenarios excluded

Never fully autonomous for:

  • Irreversible actions (delete, send, purchase)
  • High-stakes decisions (financial, legal, health)
  • Novel situations outside training distribution
  • Actions affecting third parties

AI-Native Discovery

Standard discovery practices need adaptation for AI products.

Modified Discovery Focus

Standard DiscoveryAI-Native Adaptation
"What job are you trying to do?"+ "How much do you want to delegate?"
"What's your current workflow?"+ "Which steps are you comfortable AI handling?"
"What would success look like?"+ "What errors would be unacceptable?"
"Show me how you do this today"+ "Show me how you verify AI work today"

AI-Specific Discovery Questions

Delegation appetite:

  • "Which parts of this task feel tedious vs. require your judgment?"
阅读完整来源 ↗