AI Sandbox
Hands-on program · online

From chat to AI that does the work for you. 5 live online sessions, first real use at work in 1.5-2 weeks.

ChatGPT, Copilot, and Gemini suggest. Claude Code and Codex do the whole task - you set it up and check the result.

Build your own agents and a prototype on a real case with support on Telegram. No coding required.

Olga Lunkova
Olga Lunkova

Lead Business Analyst

AI Sandbox participant

"What used to take days and require input from engineers, I now do myself with my own agents."

Agents and skills for discovery: clarifying questions, codebase checks, dependencies and blockers.

1.5-2 weeks from the start to applying AI at work

Read the full review

Before the AI Sandbox bootcamp, I felt a lot of anxiety. I read articles and kept dozens of browser tabs open, but did not know how to start applying AI in my work.

During the bootcamp, I built my first small project, published it on GitHub, and saw a tangible result I could share. Most importantly, I learned how to use AI in my work as a business analyst.

Now I use AI for almost every discovery task. I have set up agents and skills that help me ask the right questions, review the codebase, and identify dependencies and potential blockers. I can now often handle tasks myself that used to take hours and require help from colleagues.

I have also started creating templates, sharing approaches with my team, and initiating internal projects. The biggest change for me is that I no longer fear AI will replace me. Now I want to be among the people who help introduce these changes thoughtfully.

Thank you to Andrew Tomin and Sergejs Sušinskis for the practical format that helped me move from anxiety and theory to applying AI in my work.

For yourself
An open group of up to 10 people
Online, two evening sessions a week, taught in Russian. Group 7 starts on November 17, booking is open.
Price and booking →
For your team
A closed program for your company
A group from your company, shaped around the team's roles, document templates, and schedule. Invoiced to the company (B2B).
Team program (RU) →
60+
participants from real teams
25+
EPAM colleagues helped to bring AI into their projects
5
sessions of 90 minutes, the fifth is a demo day
Telegram
support and help with practical questions
Offline

Meet in person.

Open workshops and conferences featuring AI Sandbox. See all events

Program outcomes

Leave with a tool you built yourself.

You'll understand how agentic development works and start handing routine work to agents - on your own tasks, not textbook examples.

01
A project plan (PRD)
A document that spells out what you're building and why. AI drafts it with you - then you work from it.
02
A working prototype online
Not slides or mockups - a program that runs and opens by link, even on a phone. Show it to your team or in an interview.
03
Your own agents and templates
Agents and settings tuned to your tasks. Set them up once, use them every day - and build the next tools on your own.
04
The skill of checking AI's work
AI sometimes states wrong answers with total confidence. You'll learn to catch that - and rely on it calmly at work.
Who the program is for

This program is for you if

  • You're a product manager, business analyst, project manager, QA, marketer, or solo founder.
  • You've tried ChatGPT and Gemini in the browser and hit a ceiling: it's a chat, not a tool.
  • You've done certifications but are short on hands-on practice.
  • You don't write code and always assumed development wasn't for you.

This program is not for you if

  • You already build prototypes and your own agents on your own.
  • You're an experienced developer looking for deep technical detail rather than fundamentals.
  • You're expecting it to all happen automatically: here, you build with your own hands every session.
Program

5 sessions, 90 minutes each. You leave with a project online.

Evening sessions after work. There's a day between sessions for hands-on practice, so you can catch up without rushing. No programming experience required.

Session 01

From chat to tool

We install Claude Code or Codex from scratch, step by step - nobody leaves with a broken setup. Core concepts in plain language.

Result: your first working prototype on screen
Session 02

An idea becomes a plan

We turn an idea into a plan with AI: research the problem, check whether the idea is needed, and write the spec (PRD) the agents will build from.

Result: a spec (PRD) for your product or prototype
Session 03

Building your own agents

We build agents and skills on your own templates and workflows - they take over your daily routine.

Result: your own agents and skills for everyday tasks
Session 04

Landing page + MVP

We build a landing page and the first version of your product from your spec. You learn to read errors, roll back changes, and publish online.

Result: a landing page and a live MVP on the server, both reachable by link
Session 05

Demo day

Every participant shows what they built. Feedback, discussion, your plan for the next 90 days.

Result: a project demo + a 90-day plan
Participant support

Stay in touch on Telegram.

Additional rollout support
Need help putting your tools into daily work? Additional rollout support is available by separate agreement.
Support on Telegram
Stuck on an error - post in the chat and we'll work through it. No question is a silly question.
Personal consultations on request
You can arrange a personal consultation separately to review your project, questions and next steps.
Notes and recordings of every session
Written notes with visuals and a recording of each session - yours to keep. Session 1 notes are free to read before you start.
Homework and setup checklists
Before the start - a setup checklist so you arrive with a working tool. After sessions - short hands-on assignments.
The teams past participants came from
E100 EPAM Systems Box Visa Grid Dynamics Softeq Insider Digital Classifieds SEAL Group UAB ISS A/S BMC Software
Corporate format

AI Sandbox for your team.

Prototype and test a hypothesis in days, hand the routine to agents. A group of up to 10 from your company - shaped around the team's roles and departments, hypotheses, document templates, and schedule.

Corporate groups are closed, built for a company team on request. Two groups already graduated for e100.

Team program (RU) Discuss the format
01
Product teams
Prototype and test a market hypothesis in days: discovery, a spec (PRD), a prototype online, a demand test.
02
BA and project office
An agent on your templates: PRD spec, status report, risk log.
03
Operations
SOPs, request triage, reporting and reconciliation, small internal tools.
04
HR
Hiring and onboarding on agents: role profiles, criteria-based screening on anonymized resumes, the decision stays with a human.
No corporate data and no access to internal systems
Ready-made working tools for analysts and PMs stay with the team
Two instructors, demo day, certificates
Invoiced to the company (B2B)
Graduate reviews

What people who already finished say.

Real reviews from program graduates. Translated, meaning unchanged.

I'm a non-technical specialist, and this was a genuinely transformative experience. Over the course I built and shipped several working business projects: an AI agent that pulls together all calls and Slack messages and sends the client a weekly report, auto-assignment of tasks in ClickUp, a Telegram bot for TikTok content planning, and a sentiment analyzer for YouTube comments. I rebuilt the program's website from my phone mid-trip.
Anastazja Debowska
Content Strategy Consultant, ex-YouTube
LinkedIn profile ↗
My biggest takeaway: the real bottleneck in development isn't engineering capacity, it's how fast we can validate hypotheses. AI removes that barrier entirely. On the course I built an agent that takes a raw PRD draft and turns it into a finished technical spec and a Jira backlog.
Anatolii Ivantsov
Senior Business Analyst, IT & Finance
LinkedIn profile ↗
I used to think any idea needed engineers. Now, with Cursor and Claude Code, I build the first version myself, test it, and only then bring in the team. On the course I built an agent that takes a raw idea, asks the right questions, and turns it into a structured task with a spec. From idea to working prototype - no weeks of waiting for an MVP.
Sergey Khaletsky
Product Lead, SEAL Group
LinkedIn profile ↗
Working with Andrew changed how I relate to AI: not a toy chat, but a tool you actually push real work through. I used to get stuck in shallow use of GPT - a lot of motion, not much structure. Now I have my own Cursor workspace for real work: tasks, call notes, people, and projects all linked, so context doesn't reset to zero every week.
Dzmitry Pysk
Senior Operations Manager, EPAM Systems
LinkedIn profile ↗
This wasn't just talk about AI tools - we worked with them hands-on, on real examples, in a safe sandbox. Now I don't wait for someone else to set things up for me: I actively use Claude Code in my daily work, and it's completely changed my speed.
Vitaly Letyago
Program Manager / Delivery Manager, Softeq
LinkedIn profile ↗
A hands-on experience through and through: instead of talking about AI tools at a conceptual level, we worked with them directly, on real scenarios, in a guided sandbox. Andrew has a knack for turning complicated ideas into clear, ready-to-use frameworks. Most importantly, I walked away with a clear roadmap for designing AI workflows and agentic systems going forward.
Andrei Karzhou
Software Engineering Manager, Box
LinkedIn profile ↗
A rare course you walk away from not with a folder of notes but with something that actually works. Over five sessions we didn't just talk about AI tools - we built with them, hands-on, from the first idea and PRD all the way to a working prototype and deployment. Andrew has a way of breaking down something tangled into steps you can actually walk through and do. Alongside the course I started building my own thing: a web platform with a personal AI tutor for math and physics to prep for the Matura - lecture slides, solution checking, a parent dashboard. A couple of weeks ago, "I'll build it myself, no dev team" sounded like a stretch. Now I have a working prototype - plus a Claude 101 certification on top.
Andrei Rubin
Software Engineer, BMC Software
LinkedIn profile ↗
What hit hardest was the hands-on approach: not theory, but working with AI tools on real tasks. The biggest thing was how fast my thinking shifted. Now I immediately spot where automation fits into QA and which tool to reach for, and I handle most of it myself, without outside help.
Tatiana Kalinowska
Senior QA Engineer, Web & Mobile
LinkedIn profile ↗
A great course on Cursor and what AI can do. Clear, practical, and immediately applicable. What landed especially well was how complex things get explained in plain, simple language. Recommended.
Elena Truhtanova
IT Product Owner, ISS A/S
LinkedIn profile ↗
I came into the course on AI agents and Claude Code with a strong technical background already, but I still walked away with a lot that was new. The material is practical, structured, and immediately usable at work. What I valued most: Andrey guides and supports without pushing. No artificial deadlines - how much you take on is entirely up to you. This is a course for people who genuinely want to learn.
Andrei Tsiarenia
Senior QA Automation Engineer
LinkedIn profile ↗
I joined the group as a complete beginner in AI agents. Andrew walked me through the topic step by step and showed his own AI workflow - and it genuinely impressed me: agents handled everything from gathering requirements to development. Now I have a confident grasp of how AI agents actually work.
Roman Lazuko
Android Developer, Cotvec
LinkedIn profile ↗
Common doubts - answered honestly

What usually holds people back.

"This is all free on YouTube."
It is. Scattered across 40 tabs, in pieces. Here it's a single guided path with a mentor next to you, from install to a live landing page and MVP.
"90 minutes isn't enough to get anywhere."
Every session ends with a result on your screen, not a set of notes.
"I'll slow the group down."
Everyone in the group is a fellow newcomer to development: product people, analysts, QA. Nobody races ahead - we pace to the most cautious person in the room.
Who's teaching

Two practitioners.

Andrew Tomin

Andrew Tomin

Senior Product Manager, 12 years in B2B products (WorkFusion, Wargaming, EPAM, Inspectorio, Google contractor). Not a programmer - I build my own AI assistants and tools myself.

LinkedIn profile ↗
Sergejs Sušinskis

Sergejs Sušinskis

Engineering Leader, 20+ years building and leading engineering teams (CloudLinux/TuxCare, Wargaming, Profitero, AdForm). Helps take an AI idea to a working tool: debugging errors, setting up integrations, deployment, and checking results.

LinkedIn profile ↗
Enrollment

Booking group 7, starting November 17.

Group 6 for October is full. Up to 10 people per group, two evening sessions a week, taught in Russian. Corporate groups are closed, built for a company team on request.

Group 7, November
Open group · taught in Russian
Booking open
€450
This price holds until October 15, then €650 / 2850 zł.
Sessions start on November 17. Message me on Telegram and I'll hold your spot.
  • 5 live sessions, the final one a demo day
  • Your project: a spec, agents, a landing page and an MVP online
  • Telegram support, notes and recordings of every session
  • Templates, checklists, and homework

Additional rollout support and personal consultations are available by separate agreement.

Ask a question
Corporate format
Closed group
On request
A group of up to 10 from your company. Tracks, cases, and schedule shaped around the team's work.
  • Tracks: product, BA and project office, operations, HR
  • No corporate data and no access to internal systems
  • Two instructors, demo day, certificates
  • Invoiced to the company (B2B)
Team program (RU) Discuss the format
Self-paced
Start right away
€99
A self-study track built from three cohorts. No calls, no deadlines, start right after payment. Materials are in Russian.
  • 7 modules: theory, practice, prompts
  • Analyst case from the corporate cohorts
  • "From chat to agent" prompt library
  • Claude Code or Codex, your choice
About the track (RU)
Free

Useful materials

Checklists, prompts, and templates I use myself when working with AI tools. Leave your email and I'll unlock access to all of it.

See the materials
Free, no signup

Session 1 notes

The notes from session one are open online, no signup. Read them ahead of time so you arrive already oriented.

Read the notes
FAQ

Frequently asked questions.

How do I pay?
By invoice: message me on Telegram, via email, or leave a booking request - I'll send the invoice after a quick call.
Is the program tied to Claude Code?
No. We cover both Claude Code and Codex - work in either one, or try both and pick what suits you. The approaches are the same, and most of them carry over to Cursor and Google Antigravity too. What matters is understanding the mechanics: after that you move freely between tools and pick up each one's specifics yourself.
Can I bring my own project idea?
Yes. You can bring your own idea or a real use case from work - we'll work through it together so you understand how to carry it to the end. You'll leave the program with a finished solution online.
What about the security of my work data?
In the program you build your own project, not your company's real data. If you want to practice on a real work case, I'll show you how to separate out anything sensitive and what shouldn't go into a personal AI account. For the corporate format, we set up access separately, to match your company's security policy.
What if I don't have an idea of my own?
No problem. I'll give you practical examples - you can take one of our cases and build it out from start to finish.
Can I prepare before the start?
Yes, and it's worth doing. The notes from session one are open online, no signup - read them ahead of time so you arrive already oriented. And the materials include the "Claude Code in an evening" guide: the install takes about 15 minutes, so you can have everything set up before the first session. If something won't install, we'll sort it out in session one - nobody's left with a broken setup.
Which subscription do I need?
One paid subscription, for whichever tool you work in. For Claude Code the minimum is the Pro tier (from $20/mo), which covers every assignment in the program; if you work especially intensively - lots of parallel tasks, long agent sessions - it's worth upgrading to Max. For Codex you'll need a paid ChatGPT subscription. You don't need both - pick one.

Ready to build your own AI tool?

Join a group Team program (RU)

Book group 7 (starts November 17) on Telegram: @AndrewTomin.