AI Agent Builder & Operator Workshop

Become an AI Agent Builder and Operator in Two Weeks

Albert Poghosyan Founder & AI Trainer | SaaS, VR, EdTech | Techstars W17

Build AI agents that do the work, with you in control

Two Saturdays cover what takes most people months of trial and error.

Bring a task that eats your week and leave with an AI agent that handles it. No coding required. In this workshop you’ll learn to:

  • Decide which of your tasks an agent should handle, and how much freedom to give it
  • Build reusable skills connected to the apps you already use
  • Run agents on demand, on a schedule or when something happens, with guardrails that protect your data and stop prompt injection
  • Make your agents improve over time and coordinate several of them

Who it’s for:

Non-technical professionals who want to build AI agents and run them for their organization. If you already use ChatGPT and want AI to do the work rather than just answer questions, this is your next step. Bring a laptop and one repetitive task you’d like to hand off.

Dates and format:

Saturdays, 21 & 28 November 2026, 11:00–15:00

  • Two in-person Saturdays at the American University of Armenia, plus one live online Q&A a week later
  • Taught in Armenian, with templates and materials in English

Course fee:

AI Agent Builder & Operator Workshop — 75,000 AMD

Leave with a working AI agent for one of your own tasks.

Register using this link. The fee covers:

  • Both Saturdays at AUA and the live online Q&A session a week after the second Saturday
  • Ready-to-use templates: prompt templates, skills, and workflow templates you can reuse and share with your team
  • A checklist for rolling your agent out safely in your organization
  • A certificate from the American University of Armenia

What you’ll learn

Design, build, secure, and improve an agent for your own work. You’ll build hands-on with Claude, ChatGPT, n8n, Make, Gumloop, and Gemini, and learn which one fits your organization. Every exercise is built around your own work: before day one, our system asks about your role and tasks and generates your practice exercises.

Decide what to delegate
  • Sort your tasks by how often they happen, what starts them and how much review they need
  • Choose whether an agent should remind you, advise you, draft for your approval, or act on its own
  • Give it more freedom only as it proves reliable
Build reusable skills
  • Turn a repetitive task into a skill with clear steps, an output format and limits
  • Connect it to the apps you already use, such as email, calendar, Drive, Notion or your CRM, through connectors and MCP
Run agents without babysitting (bot-sitting)
  • Start them on demand or on a schedule
  • Trigger them when something happens, such as a new email or a submitted form
Set boundaries before you grant access
  • Control which files, accounts and websites an agent can reach, and require your approval for sensitive actions
  • Protect passwords and personal data, in line with Armenia’s personal data law
  • See how instructions hidden in emails and web pages (prompt injection) can hijack an agent, and how to stop them
Make agents improve over time
  • Define what a good result looks like
  • Log your corrections and catch repeated mistakes
  • Update your skills, approving every change yourself
Coordinate several agents and control costs
  • Know when to split work across agents, and when a smaller model is enough
  • Give agents only the context they need from a shared knowledge base

Workshop agenda

At least 70% of every Saturday is hands-on building.

  1. [Before day one] Setup

    Follow a short setup guide to create your accounts, pick your build tool, and choose one real task to delegate.

  2. [Saturday, 21 November] From chatbot to agent (11:00–12:30)

    What agents can and can’t do; sort your own tasks and pick the first one to hand off.

  3. Break (30 minutes)

    Lunch and questions.

  4. [Hands-on] Build your first skill (13:00–15:00)

    Write the skill, connect it to your apps and run it on sample data.

  5. [Saturday, 28 November] Safe, self-improving agents (11:00–12:30)

    Permissions, approvals and prompt injection; feedback loops; schedules and triggers; handoffs between agents.

  6. Break (30 minutes)

    Lunch and questions.

  7. [Hands-on] Finish, test and demo (13:00–15:00)

    Add guardrails and a review loop, test your agent and show it to the room.

  8. [Online] Live Q&A (one week later)

    Bring what broke in your first week of real use and fix it together.

Learn directly from Albert

Albert Poghosyan Contact

Albert Poghosyan

Founder & AI Trainer | SaaS, VR, EdTech | Techstars W17

Albert Poghosyan has spent 14 years building and scaling companies in SaaS, VR and EdTech. In 2016 he took a VR startup from $0 to $3M in annual recurring revenue in 14 months by focusing on enterprise sales. His companies went through Techstars (W17), Plug and Play and NVIDIA Inception, and were backed by Quake Capital and HTC Vive.

He now trains professionals to put AI agents to work, including the Project Management with AI executive masterclass in Yerevan.

Accelerators & investors

  • Techstars
  • Plug and Play
  • NVIDIA Inception
  • Quake Capital
  • HTC Vive

Comments are closed, but trackbacks and pingbacks are open.