Before
You start in a blank chat and keep adding context after AI gets the task wrong.
After
You give AI the goal, context and limits before it starts.
AI1: Six live sessions in Czech
Make AI Do What You Need
Stop re-explaining tasks to AI and fixing what it gets wrong. Over six weeks, build a way to give AI useful work, check the result and get the job done.
- Work on a task you need to finish.
- Learn how to check AI's work.
Make AI Do What You Need is AI1 of the AI Program: six live online sessions in Czech for engineering managers and senior engineers who want AI to finish real work. Bring work you already need to do. Use what you learn between sessions, then get direct feedback on what worked and what did not.
6 Monday sessions, 09:00–12:00, 19 Oct – 23 Nov 2026. Online in Czech.Led by Petr Glaser. Cohort capped at 12, runs from 6. Closes when full.
You pay for yourself: 64,000 CZK per seat. Your company pays: 98,000 CZK per seat
Trusted by
Not for
- People who want to roll AI out across several teams. That is AI2: Spread the AI, the next level.
- Engineers who already run agents in production and want peers, not lessons. That is AI Masters.
- Anyone looking for a prompting cheat sheet. You build things and run them in each session.
Who it is for
- Engineering managers and team leads whose teams already use AI. Learn by doing so you can review the work and lead it.
- Engineering directors who need to understand how the work gets done, even if they no longer write code every day.
- Senior and Staff engineers who use AI often but still spend too long retrying prompts and checking the output.
Get the whole job done with AI.
- Grill the task
- Map the code
- Design and split
- Run agents
- Review and test
- Hand it over
- Save as skills
AI can produce a first draft in minutes. Review, tests and handoff still take time. You will work through the whole job on your own tasks, then bring the results back each week.
The program
Six sessions on real work
Grill the task before AI touches it.
Bring a vague task from your own work. Before any code, the agent asks about missing information, risks, limits and what done means, and you mark the decisions only a person can make. You leave with a short brief, the open questions and your first checkpoints.
Map unfamiliar code or a PR before you change it.
Point the agent at a codebase or PR you barely know. Trace the entry points, calls, dependencies, tests and what a change would affect, and keep what the code proves apart from what the agent guesses. You leave with a short map of the affected part and context you have checked.
Design the change and split it into safe tasks.
Compare options and cut the one you pick into slices you can verify. Draw a DAG, a map of what depends on what, and turn each slice into a sub-issue with inputs, outputs, boundaries and a stop condition. You leave with a plan and a first queue of tasks you can safely hand to agents.
Run agents on a bounded job and fix what breaks.
Give an agent one task you can check, and run agents in parallel only when the work is independent. Read the output as it comes, deal with failures and pick the model and cost that fit each kind of task. You leave with your first shipped change and a log of what you decided and what went wrong.
Prove the result holds.
Have an agent review the change from a different angle than the agent that built it, and cover the critical properties with tests, types or other deterministic checks. Then see the change through debugging, review, testing and handoff until it is done. You leave with checks tied to your risks and a change you can defend.
Turn what worked into your own system.
After every session you capture one step that worked as a small skill. Now you test each one on new input, cut it down and combine the set into one personal workflow, then run it on a second real task. You leave with a set of reusable skills and two real tasks where you used them.
What changes after six weeks
Before
Your weekly report or incident summary starts from scratch every time.
After
You reuse a skill you built for that task.
Before
AI gives you a plausible answer, but you still have to check every part of it.
After
You know which parts need your judgment and which checks you can automate.
Before
The first draft is fast. Review, tests and handoff still hold the work up.
After
You plan for those steps and keep the context the next person needs.
What you can do by the end
- Turn a vague task into a clear brief before AI writes anything.
- Find your way in unfamiliar code or a PR and check what the agent claims.
- Split a change into small tasks and run agents on them with a clear stop condition.
- Prove the result with review and checks before it reaches your team or customer.
- Reuse your own skills on a second task without starting over.
Format
Format and price.
- Six Mondays, 09:00–12:00 Prague time, from 19 October to 23 November 2026. Taught online in Czech. We use Claude for exercises. Bring your current AI tools; we will check access with you before the first session.
- Petr Glaser leads the sessions. Marian Kamenistak opens and closes the cohort.
- 12 places. We run the course with at least 6 people. If fewer than 6 have joined by 9 October, you get a full refund.
- Pay when you reserve. After session 2, you rate the course in a two-question form. If the average is under 7/10, you get a full refund, no questions asked.
You pay for yourself64,000 CZK per seat
Your company pays98,000 CZK per seat
You pay for yourself: Invoice in your name. One or two seats. Your company pays: Invoice to the company with its VAT ID, entered at checkout. Up to 12 seats in one booking.
Register and pay within three days after the webinar and you also get a 1:1 call with Petr about your own AI use case.
Refund rule
- When you pay
- On reservation, via Stripe
- Who scores it
- You, after session 2, in a two-question form
- Average under 7 out of 10
- Full refund, no questions
- Fewer than 6 seats by 9 Oct
- The cohort does not run, full refund
Cohort capped at 12, runs from 6. Closes when full.
Notify me about the next cohort
Prices in CZK, excl. VAT. Paid on reservation via Stripe.
What nobody else offers
Why this course helps
01
Work on your own tasks.
Bring a task you need to finish. Petr helps you apply the method to it.
02
Use it the same week.
Try each step at work before the next session. Bring back the result and get feedback.
03
Learn to check the output.
Build reviews and checks around the mistakes that matter in your work.
04
Learn alongside other leaders and engineers.
The group is small enough to discuss real work, not only watch a demo.
“Review and testing decide whether faster coding saves you any time.”
Proof, honestly labeled
Evidence from earlier Academy courses
- This course is part of the Engineering Leaders community, with 3,300+ members across Prague, Brno, Bratislava and Krakow.
- Companies that trained their managers with us: Rohlík, Košík, Mews, Kiwi.com, Česká spořitelna, Trezor, and Apify.
- 9.3/10 average from 21 participant reviews across the Academy programs this one grew out of.
This is the first AI1 cohort. The ratings and company logos come from earlier Academy programs, not this course.
Questions
Questions about AI1
Do I need to be technical?
You do not need to code. Some exercises use a terminal, and Petr will walk you through them.
Which tools do we use?
We use Claude for shared exercises. Bring the AI tools you already use; before session 1 we confirm what fits and what access you need.
What happens if I miss a Monday?
You get the recording and can catch up before the next session.
What if the cohort does not fill?
It runs from 6 people. We decide on 9 October and refund you in full if it does not run.
Can my company pay for the seat?
Yes. Enter your company details at checkout and Stripe issues the invoice.
Where to go next.
More from the Academy
In the ELC community
Make AI Do What You Need
Bring a real task. Leave knowing how to do it with AI and check the result.
Ready for the full operating rebuild? The AI-Native Engineering Manager →