Agentic Does Not Mean Fully Autonomous | Agentic AI Explained

Ask an AI to test an application and find anything important that is wrong. A typical AI assistant might write you a list of test cases. An AI agent might run those tests. An agentic AI system works out what the goal means, decides what to do, uses tools, looks at what it finds, changes its approach when it needs to, and keeps working towards the goal. Agentic AI means AI systems designed to work towards a goal by planning, making decisions, using tools and acting on an environment, with some or all of those decisions made on their own. It works as a loop: goal, plan, act, observe, evaluate, adapt. Every turn of that loop is a decision, and every decision is a point where the system can go wrong, so each one needs checking. A common belief is that agentic AI means fully autonomous AI. In practice, many agentic systems keep a person in the loop: an agent might run tests on its own but need approval before it changes production data. The level of autonomy is a design decision, and you test that the limits hold. The episode also follows an agentic testing system through a checkout failure investigation, covers the four places agentic AI changes testing, the six questions to ask about the agent itself, and how assistant, agent and agentic differ. ⏱️ Chapters 0:00 Three answers to one request 0:27 What the AI Tester Dictionary is 0:38 Working towards a goal 1:15 A loop of six steps 1:43 More than one step 2:24 Levels of autonomy 2:56 Investigating checkout failures 3:49 Four places testing changes 4:21 The agent is under test too 4:43 Look at what it does 5:19 A person in the loop 5:48 What to hold on to 🔑 Key takeaways • Agentic AI describes systems that work towards a goal through a loop of planning, acting, observing, evaluating and adapting • An agentic system is more than a model: it combines the model with tools, memory, instructions, an environment, feedback and control, so test those parts as well as the model • Agentic does not mean fully autonomous: the level of autonomy is designed, so check the system stays inside its limits and asks for approval when it should • Before you trust an agent's report, open the evidence and check that it supports the conclusion • You now test the agent as well as the application: its plan, its decisions, its tool calls, its permissions, its recovery, and whether it reached the goal 📚 In this series #001 — Artificial Intelligence #002 — Machine Learning #003 — Deep Learning #004 — Generative AI #005 — Large Language Models #006 — Transformer #007 — Diffusion Model #008 — Foundation Model #009 — Token #010 — Context Window #011 — Prompt #012 — System Prompt #013 — Prompt Engineering #014 — Temperature #015 — Top K and Top P #016 — Hallucination #017 — Embeddings #018 — Vector Database #019 — Semantic Search #020 — Retrieval-Augmented Generation #021 — Chunking #022 — Reranking #023 — Knowledge Graph #024 — Multimodal AI #025 — Function Calling #026 — Structured Output #027 — Grounding #028 — Fine-Tuning #029 — Context Engineering #030 — AI Agents #031 — Tool Calling #032 — Agentic AI ← you are here ▶️ Previously #031 — Tool Calling: how an AI model asks the application to run a tool, and what to check on every call.    • The AI Can't Run Your Tests. It Asks for a...   ▶️ Next in the series #033 — Multi-Agent Systems: several agents working together on one goal. 🔔 Subscribe for practical AI and software testing content. #SoftwareTesting #AIForTesting #AITesterDictionary #AgenticAI #AIAgents #AutonomousAgents #QAEngineer #SDET #AITesting #TestAutomation #HumanInTheLoop #QualityEngineering #SoftwareQuality FOLLOW ME 📸 Web: https://softwaretestingtrends.com/ Discord:   / discord   LinkedIn:   / ambysan   Twitter:   / ambysan   Facebook:   / softwaretestingtrend   Medium:   / ambysan   BLOGGING 📝 Software Testing Trends: https://softwaretestingtrends.com/ TechJournal: https://ambreenkhan.dev/ ABOUT ME 👩‍💻 https://ambreenkhan.netlify.app/