AI Agents & Automation: Build Your Own AI Workflows [EN]

This course contains the use of artificial intelligence.

You have been using AI for a year. It answers well. It has never once done anything while you were away from the keyboard.

Where the gap is

Between writing a good prompt and having something that runs on its own there are four steps, and almost everyone stops at the first. You get a useful answer, paste it somewhere, and do the same thing again next week. The vocabulary does not help: agent, workflow, MCP, skill, plugin and orchestration all get used interchangeably by people selling things, so it is genuinely unclear what you are supposed to build, in what order, or whether any of it needs a developer. It does not. But it does need the four steps in sequence, and the gap between them is where the time goes.

What this course covers

Thirty-three lessons through all four. The prompt as a construction first: seven blocks including the validation block most people skip, four classic frameworks, chain-of-thought and self-consistency, tree-of-thoughts for multi-criteria decisions, ReAct where the model chooses its own steps, then the shift to context engineering — projects, custom instructions, retrieval without a developer, and memory across sessions. Then what agents currently do: the components of one, the platforms, real products across six categories with their limitations stated, and building one from scratch. Then the platform layer in depth: where projects, skills, plugins and connectors sit, writing SKILL file, adapting official commands, and packing everything you built into a single distributable plugin with a manifest. Then production: auditing your own week to find the leverage, the augment-automate-architect cycle, data boundaries, building a skill live, connector security, scheduled tasks, cross-application workflows, Computer Use, and proving the time saved. Finally code-level agents: browser work, working across your applications, report agents, unattended automations with checks and rollbacks, and orchestrating five agents with context handoff and an audit log.

A note on the examples

The worked examples come from a people-function context — screening a stack of CVs, building an onboarding pack, assembling a weekly report. That is deliberate: it is a domain with enough repetitive document work to demonstrate a full cycle from prompt to plugin. Nothing about the techniques is specific to it. Seven-block prompts, tree-of-thoughts, MCP, SKILL file, n8n, scheduled tasks and multi-agent handoff behave identically whatever the task is. Also worth knowing: this course names specific products and versions, and some of those will look different in a year. The principles underneath — prompt architecture, the prompt-workflow-agent distinction, human in the loop, data boundaries — will not.

Who is teaching this

I am Mike Pritula. I built the people system at Preply as it became a unicorn, and I have worked at Wargaming, iDeals and Alfa-Bank. More than 1.6 million students have enrolled in my courses across 185 countries, and over 150,000 specialists have gone through my programmes. I hold PHRi and SHRM-CP certifications and represent HRCI in more than ten countries.

What is included

• Lifetime access to all 33 lessons

• Active instructor support in the Q&A section

• A Udemy Certificate of Completion

• Working material: the seven-block prompt, four prompt frameworks, SKILL file templates, connector safety checklist, scheduled automation patterns, the multi-agent handoff model, a 90-day roadmap

• Everything built without writing code

Where to start

Name one thing you do every week that produces a document from other documents. That is the task this course automates, and by the end it should be running on a schedule without you. Enrol now and start today.