Agentic AI & AI Agents Development

Move from individual prompts to an AI system that can work through a defined task using permitted tools. Learn when a predictable workflow is enough, when an agent is useful and how to keep actions observable and controlled. Projects use fictional data and a practice environment.

What you will learn

• Design a bounded agent with clear goals, inputs, tools and completion criteria.
• Connect tools and structured data while managing state and failures.
• Add approval steps, access limits, budget controls and useful logs.
• Evaluate a workflow against test cases instead of judging only a successful demo.

Planned course syllabus

1. Agents versus workflows: Understand planning, tool use and state; choose the simplest design that meets the task.
2. Task and tool design: Define tool inputs, structured outputs, permissions and allowed actions for a focused assistant.
3. APIs and orchestration: Connect practical steps in a visual workflow or small coded prototype; handle timeouts and errors.
4. Context, retrieval and memory: Use supplied knowledge and task state carefully; define what should be retained and what should be discarded.
5. Planning and controlled execution: Break tasks into steps, validate intermediate outputs and require human approval for consequential actions.
6. Coordinating multiple roles: Compare one-agent and multi-agent designs, handoffs, duplication and failure propagation without adding needless complexity.
7. Testing, security and cost: Measure task success, reliability and tool use; test prompt injection, excessive permissions, loops and budget limits.
8. Final supervised agent: Build a bounded agent workflow, demonstrate normal and failure cases and document operation and maintenance.

Practical assignments

Map a workflow, define a tool contract, add an approval gate and test recovery from a tool failure. Compare a fixed workflow with an agent approach for the same task.

Final project

A supervised document-to-action assistant using fictional inputs: extract information, prepare a structured result and propose a next step. Include at least two tools, human approval before external actions, logs, retry limits and a test report covering ten scenarios.

Who this course is for

Learners with AI prompting and basic workflow experience; introductory API or coding familiarity recommended.

Tools and learning approach

Practice can use a visual workflow builder and introductory API or code examples. Basic prompting is expected; API and programming foundations are recommended. Production credentials, paid services and real business actions are not required for the core project. Free-tool options will be used where practical. Optional paid software, subscriptions or API usage are not included and will be identified before enrollment.

Course availability

Coming soon. This page presents the planned syllabus and project outline. Lesson materials, instructor details, fee, duration and class schedule will be confirmed before enrollment opens.