AI Agents for Professionals
A hands-on micromasters program for non-technical professionals who want to move from chaotic AI experiments to systematic automation of their work processes.
overview
You already use ChatGPT, Claude, Gemini or Copilot, but the routine is still on you. AI can write a draft, summarise a meeting or help with an idea. In real work, though, that quickly stops being enough: the context has to be explained all over again, the output has to be checked constantly, and turning AI work into a system looks like something vast and hard to grasp.
AI Agents for Professionals helps you put all of this into a system. Over 8 weeks you will break down your own work process, build AI agents for real tasks, connect them to your working data through MCP, check their quality and prepare a capstone with a measurable effect for yourself, your team or your manager. A separate value of the program is 60 hours of expert support from Jane Klepa: feedback on assignments, review of interim results, consultations on the capstone project and final defences.
what you will learn
requirements for participants
Program
Workshop 1. From a chat prompt to an agent
- You understand what an AI agent is and identify 3 of your own tasks worth testing for agent automation
- agentic AI and how it differs from chat-based AI
- basic security
- the landscape of tools
- the logic of an agent: instructions, context, tools, autonomy
- Audit your own work processes and pick 3 candidates for agent automation
Workshop 2. Claude Projects as your workspace
- Your first working agent for your own task
- custom instructions, project knowledge, files
- instructions for a predictable result
- context management
- Create your first Claude Project for one of your tasks and test it on 10 real cases
Workshop 3. MCP: the agent goes beyond the chat
- Your agent works with your real working data
- what MCP is
- connecting to Gmail, Google Drive, Notion, calendars, databases
- ready-made MCP connectors and the situations where you need your own
- Connect at least 2 MCP services to your agent and complete a real task end-to-end
Workshop 4. Designing an agent for a real task
- Your agent is built on a clear workflow map rather than on intuition
- workflow mapping for AI agents
- steps, dependencies and control points
- anti-patterns
- the instruction → test → fix → test loop
- Refactor your agent on the basis of the workflow map
Workshop 5. Evaluating quality and improving the agent
- You can show that your agent works well, not just that it seems fine
- the quality of an agent’s work
- metrics for non-technical tasks
- checking the output regularly
- human-in-the-loop
- Build a mini evaluation set of 5–8 cases and run your agent through it
Workshop 6. Multi-step automations: Make, Zapier, n8n
- Your agent is built into a workflow with a trigger, actions and execution logic
- when one agent is not enough
- Make, Zapier and n8n
- triggers, actions, branching, AI steps
- choosing a tool by task, budget, complexity and data requirements
- Choose a tool for your task and build one automated workflow with a trigger
Workshop 7. Security, data, ethics
- You understand the limits of trust in an agent and have usage rules for yourself or your team
- which data must never be passed to an agent
- confidentiality and compliance
- prompt injection
- labelling the use of AI
- the points where control returns to a human
- Audit your agent for risks and prepare usage rules
Workshop 8. Capstone presentation and scaling
- A final presentation of your AI solution and a plan for developing your own AI practice
- capstone project presentations
- how to present the solution to your team or your manager
- how to describe the effect: time, quality, scale, risks
- what to do with your agent after the program ends
- Defend your final project and get feedback
Capstone project
Throughout the program you work on your own AI agent or system of agents for a real work task. At the end you present the task to be automated, the architecture of the solution, a demo on 3 real cases, the measured effect, the risks and the usage rules.
Evaluation criteria: functionality, how realistic the task is, the quality of the agent design, the clarity of the presentation, the risk assessment.
Instructor
Jane Klepa
Director of Innovation and Entrepreneurship at SET University An entrepreneur, former Executive Director of 1991 Accelerator, project coordinator, mentor and speaker with 12 years of experience in technology. Jane also teaches the “Developing entrepreneurship mindset” course on SET University’s master’s programs.
Benefits
Expert support — 60 hours set aside for feedback, reviewing assignments, consultations on capstone projects and final defences with Jane Klepa
Hands-on — you work with your own real process, the one you can start automating during the program
No-code — no programming, but with the logic of agent design, workflow mapping and quality checks
Tool-first — Claude Projects, Copilot, MCP, Make, Zapier or n8n on practical tasks, not in theory
Business impact — a focus on how to show the result through time, quality, budget, ROI or less routine
Security and control — working with data, compliance, prompt injection and human-in-the-loop, without the illusion that AI can simply be “let loose”
Practical result — an AI agent or workflow for your own task, an evaluation set and a final capstone presentation
who it is for
Middle and senior non-technical professionals who already use AI at a basic level and want to embed it into their work systematically
Marketing, communications, sales, business development, HR, L&D, operations, finance, legal, product, project management, analytics, support and delivery teams
The program is for those who need a result they can show to a manager or a team: the process, the effect, the risks and the room to scale
FAQ
Do I need programming skills?
No. The program is no-code. If you are looking for a course on building agents with LangChain or LangGraph, this program is not about that.
I already use ChatGPT every day. Will it still be useful for me?
Yes. The program is built for people who are past the stage of basic prompts and want to build AI processes systematically: with context, integrations, quality evaluation and business impact.
What if my company does not have an AI strategy yet?
That is a normal situation. The program helps you start with the processes you really influence in your own role and show the result in the language of business.
Do I have to use real work data?
You work with your own real task, but you can adapt the data so that you do not break your company’s internal rules. Confidentiality, compliance and the limits of AI use are covered separately during the program.
How much time does it take per week?
6–8 hours: 2 hours of live workshop and 4–6 hours of self-paced work on your own project.
Do I need paid versions of the tools?
Yes. Claude Pro is required, roughly from $20 per month. You will also need one automation tool of your choice. The estimated budget for tools over the program is $40–60.
What will I have at the end?
A working AI agent or workflow for your own task, an evaluation set, a risk assessment and a capstone presentation with a measurable effect.
Build your own team of AI assistants and turn daily routine into a workflow you control
