SmartGT Academy · Cohort 03 · starts 1 December

Claude Code Operator

A new role companies are opening right now: a person who runs an AI agent and is accountable for the result.

$500 · UAH 18,000 · withdraw after the first session

  • 6weeks
  • 12sessions of 90 min
  • up to 8people per group
  • +1month of support
Claude Codecourse tool
Task

Add report export to XLSX.

Scope: the reports module only.

Done when the tests pass

and the file opens in Excel.

Plan

1. Find the report builder

2. Add an exporter

3. Write a test

→ approved by the operator

Diff

reports/export.ts+42 −3

reports/export.test.ts+28

Acceptance

tests 18/18 · manual check

→ accepted

One operator task: from setting it to accepting it.

Who the course is for

You don’t have to code. You do have to know your field.

  • Developers and team leads

  • Analysts and system administrators

  • Team managers

  • Document workflow and lawyers

What you need
  • A laptop with access to a work project
  • Terminal and git at the level of clone, branch, commit
  • A Claude subscription with access to Claude Code — we’ll help you set it up

What an operator can do after the course

01

Holds the context

Describes the project so the agent doesn’t ask the same questions every week, and knows what must never go into the context.

  • CLAUDE.md and project memory
  • Access boundaries for code and data
02

Breaks down the task

Gets a plan before any code appears and stops work at a wrong hypothesis, not after it.

  • Plan before changes
  • Definition of done
03

Accepts the result

Reads the diff, not the summary. Knows which checks to run before saying “done”.

  • Line-by-line review of changes
  • Tests and a manual check
04

Builds the process

Turns a one-off task into a tool the whole team uses, not just the operator.

  • Custom skills and commands
  • Procedures for working with the agent

Programme: eight modules

Every module ends with practice on your own task.

M01

The operator’s workstation

Installation and working modes, permissions and sandbox, model choice, plan mode, reading changes in the editor.

M02

Project context

Project description, memory, repository structure. What the agent sees and doesn’t see — and why “it doesn’t know our code” is almost always the operator’s mistake.

M03

Task setting

From “make it look nice” to an acceptable task: scope of changes, definition of done, plan before code, stopping a wrong direction.

M04

Accepting the work

Reviewing changes, tests and checks, typical places where the agent fails silently. What to do when the result is almost right.

M05

Skills, commands, subagents

How to turn repetitive work into a tool: a custom skill, a command, a subagent for parallel search. And when a subagent is overkill.

M06

Connectors and integrations

Connecting Microsoft 365, Google and your own services: an agent that works with email, calendars and accounting data, not just files.

M07

Security and boundaries

Secrets, personal data, client isolation, permissions and action interception. What must never be handed to the agent and where the operator’s responsibility ends.

M08

Team rollout

Branches, checks, code review, automated runs. How to measure the impact and what to do about resistance within the team.

Starts 1 December · groups of up to 8 people

Join cohort 03

Cohort trainer

Mykhailo Vasianovych, founder of SmartGT

Mykhailo Vasianovych

Founder of SmartGT. Teaches the course and assesses operators.

“An agent doesn’t replace an engineer. It changes what an engineer is paid for: not lines of code, but a well-set task and an accepted result. That is exactly what we teach.”
  • since 2012in business digitalisation
  • 4own products

    Smart 360 · Smart LMS · Smart CMS · BOT-panel

  • 50+client websites on our own platform
  • dailyworks with Claude Code

Course practice uses tasks from our own products.

How the training works

90 min your task 90 min
  • Wk 1
  • Wk 2
  • Wk 3
  • Wk 4
  • Wk 5
  • Wk 6
  • Online, in Ukrainian
  • 2 sessions a week, 90 min each
  • Everyone shares their screen
  • You keep the recordings

The course is the start. Then comes the rank

Ten levels. Each is confirmed by hours, sessions, tokens and a practical assignment.

R0 0 Connected
first conversation

Tool installed, permissions configured.

R1 1 Intern
6 h · 3 sessions 10M tokens

Handles supervised tasks. Stops the agent when it heads the wrong way.

R2 2 Operator
14 h · 7 sessions 25M tokens

Runs a task alone: from setting it to an accepted diff.

R3 3 Reviewer
26 h · 13 sessions 45M tokens

Reviews the agent’s work done by others. Catches errors before merge.

R4 4 Skill Master
42 h · 21 sessions 75M tokens

Captures the department’s routines in the team’s tools.

R5 5 Integrator
60 h · 30 sessions 110M tokens

Connects the agent to email, calendars and accounting systems.

R6 6 Warden
80 h · 40 sessions 145M tokens

Responsible for access, secrets and isolation of client data.

R7 7 Conductor
95 h · 48 sessions 175M tokens

Runs several agents in parallel on one product.

R8 8 Architect
180 h · 90 sessions 330M tokens

Owns the process in the department: branches, checks, CI, procedures.

R9 9 Mentor
350 h · 175 sessions 640M tokens

Trains operators and upholds the standard for working with the agent across the company.

  • During the courseThe first two or three levels are earned through practice on your own tasks.
  • After the courseA monthly pulse and assessment for the next rank.
  • In the companyThe rank is a clear benchmark for hiring and role reviews.

Cohort 03 — starts 1 December

  • You keep the session recordings
  • Support chat for a month after the course
  • Payment under a contract; withdraw after the first session

Questions before you enrol

I’m not a programmer. Will there be anything for me to do on the course?

Yes. Half of the practice isn’t code: procedures, reports, checking documents against a checklist, working with email and calendars. The only requirement is basic terminal skills, and we teach that in the first module.

Do I need my own Claude subscription?

Yes, the course runs in your own working environment. If you don’t have a subscription yet, we’ll help you choose a plan and set it up before the cohort starts.

Can we give the agent access to our work data?

That’s the subject of a dedicated module. We cover what must never be handed to the agent, how permissions and action interception work, and what an access policy looks like that you can confidently show your security team.

We want to train a whole team at once.

A corporate group of up to eight people runs as a separate cohort on dates that suit you, with practice built on your tasks. Get in touch and we’ll tailor the programme to your department.

What happens after the course?

A support chat for a month and a monthly operator survey: we ask what you’ve managed to build into your processes and what’s getting in the way. The combined answers shape future cohorts.

SmartGT is an IT team providing comprehensive business solutions. SmartGT Academy is an independent educational project. Claude and Claude Code are trademarks of Anthropic; we are not an official partner of the company.