Belajar Quality Engineer - Agentic AI Testing

#Pendahuluan

Setelah di episode 21 kita mempelajari AI-assisted quality engineering, pada episode ini kita mempelajari agentic AI testing — bagaimana menggunakan AI agents untuk testing dan verification. Agentic AI adalah evolusi dari AI-assisted: AI agents bisa menjalankan testing secara autonomous dengan minimal human intervention.

Mengapa agentic AI testing penting? Karena AI agents bisa mengeksekusi testing secara continuous, menemukan bugs yang mungkin terlewat oleh manual testing, dan memberikan feedback secara real-time. Ini memungkinkan QE untuk fokus pada strategi dan improvement.

#Konsep Agentic AI

#Perbedaan AI-Assisted vs Agentic

text

AI-Assisted vs Agentic:
├── AI-Assisted:
│   ├── Human executes, AI assists
│   ├── AI memberikan suggestions
│   ├── Human做出 keputusan
│   └── Semi-automated
└── Agentic:
    ├── AI executes autonomously
│   ├── AI做出 keputusan
│   ├── Human监督 & approve
│   └── Fully automated

#AI Agent Capabilities

text

AI Agent Capabilities:
├── Planning:
│   ├── Break down tasks
│   ├── Create execution plan
│   ├── Prioritize activities
│   └── Allocate resources
├── Execution:
│   ├── Run tests
│   ├── Collect results
│   ├── Analyze outcomes
│   └── Report findings
├── Learning:
│   ├── Learn from results
│   ├── Adapt strategies
│   ├── Improve accuracy
│   └── Build knowledge
└── Collaboration:
    ├── Communicate with humans
    ├── Coordinate with other agents
    ├── Escalate issues
    └── Request approval

#Agentic Testing Workflow

#Autonomous Test Execution

text

Agentic Testing Workflow:
├── 1. Task Assignment:
│   ├── Human defines scope
│   ├── AI agent accepts task
│   └── AI creates execution plan
├── 2. Test Design:
│   ├── AI generates test cases
│   ├── AI creates test data
│   └── AI prioritizes by risk
├── 3. Test Execution:
│   ├── AI runs tests
│   ├── AI collects results
│   └── AI handles errors
├── 4. Result Analysis:
│   ├── AI analyzes failures
│   ├── AI identifies root cause
│   └── AI suggests fixes
└── 5. Reporting:
    ├── AI generates report
    ├── AI highlights critical issues
    └── AI recommends next steps

#Tools & Platforms

text

Agentic Testing Tools:
├── Modern AI Testing:
│   ├── Testim (AI-powered)
│   ├── Mabl (intelligent automation)
│   ├── Functionize (AI test creation)
│   └── Applitools (visual AI)
├── AI Agent Frameworks:
│   ├── LangChain
│   ├── AutoGPT
│   ├── BabyAGI
│   └── Custom agents
└── Platforms:
    ├── GitHub Copilot
    ├── Cursor
    ├── Codeium
    └── Custom AI platforms

Note

Agentic AI testing masih dalam tahap awal. Mulai dengan AI-assisted testing dan bertahap adopt agentic capabilities saat tools dan platforms matang.

#Practical: Agentic Testing Plan

yaml

agentic_testing_plan:
  phase_1:
    name: "AI-Assisted"
    scope: "Test generation & triage"
    automation: "50%"
    human_involvement: "High"
  phase_2:
    name: "Semi-Agentic"
    scope: "Test execution & analysis"
    automation: "70%"
    human_involvement: "Medium"
  phase_3:
    name: "Agentic"
    scope: "Autonomous testing"
    automation: "90%"
    human_involvement: "Low"
  tools:
    - ai_assisted: "Testim, Mabl"
    - agentic: "Custom AI agents"
    - monitoring: "Observability platform"

Tip

Mulai dengan AI-assisted testing untuk test generation dan triage. Setelah terbiasa, tambah semi-agentic capabilities untuk test execution. Bertahap lebih baik daripada big bang.

#Penutup

Pada episode 22 ini, kalian telah mempelajari agentic AI testing.

Inti yang harus dibawa pulang:

Agentic AI: AI agents menjalankan testing secara autonomous.
Capabilities: planning, execution, learning, collaboration.
Workflow: task assignment → test design → execution → analysis → reporting.
Adoption: mulai AI-assisted, bertahap ke agentic.

Di episode 23 selanjutnya, kita akan membahas testing AI/LLM systems — bagaimana menguji sistem AI/LLM secara komprehensif. Sampai jumpa di episode 23!