AI training for software and IT teams: not the team that writes the code, the team that manages it
Full day (about 7 hours) Up to 10 people On site in Istanbul or online 49.900 TL + VAT
Software and IT
What is this program?
AI training for software and IT teams is a full-day, hands-on program in which development, test and systems teams learn to set up code generation, test writing, debugging, documentation, code review and agent-based development flows on their own codebases with tools such as Claude Code, OpenAI Codex and GitHub Copilot.
In 2026 the question for software teams is no longer "can AI write code" but "how should the team work with AI". Claude Code, Codex and Copilot are no longer line-completion tools. They are agents that take a task, move between files, run tests and make fixes. The speed gap between developers who use these tools and those who do not is turning into tension inside teams.
The program brings the whole team onto the same working model: how a task is described, within which limits the agent works, how generated code is reviewed, where testing and security stand. Throughout the day the team works in its own repository on a real task.
Who is it for?
Software development teams
Teams building web, mobile or back-end systems with a heavy code review and testing load.
IT and systems teams
Teams that write scripts, maintain system documentation and have recurring operational work.
Technical managers and product teams
Those who will decide under which rules the team uses AI tools.
Program flow: six modules
Talking takes at most a fifth of the day. The rest is hands-on production on your own files, each module ending with a piece of real work.
Tool map and working model
The real differences between Claude Code, Codex, Copilot and Antigravity. Line completion, chat and agent modes. Which work is done in which mode.
Task description and context management
How a task is described to an agent: scope, constraints, acceptance criteria. Writing the project context file. With the team's own repository.
Code generation, testing and debugging
The loop of feature development, generating tests, finding and fixing bugs. Rules for reading and rejecting generated code.
Code review and documentation
AI-assisted code review, change summaries, generating API and system documentation from code.
Security, privacy and quality rules
Which code and data may enter a tool, secret key risk, licensing and copyright, a quality gate for generated code.
Agent-based development and the team standard
Handing multi-step tasks to an agent, approval points, cost ceilings. The team's written AI working standard.
Real work scenarios
The kind of tasks the day is built around. Yours replace these after the discovery call.
A pending feature request
Describing a real task to an agent, developing it end to end and having the team review it.
A module with low test coverage
Writing and running a test suite for an existing module with AI.
Documenting legacy code
Documenting an undocumented service by reading it from the code.
A production bug
Going from the error log to the root cause and evaluating the fix suggestion as a team.
A recurring IT script
Turning a manual system task into a script and checking it for security.
Team working standard
The team's own written rules for AI tools: what it uses, what it does not, how it reviews.
Tools used
Skills come before tools. The current model comparison with prices is on the AI models page.
What the team leaves with
- A project context file written for the team's own repository
- A working model with task description, acceptance criteria and review rules
- Ready flows for testing, documentation and debugging
- A written security rule for secret keys, licences and data
- Approval points and a cost rule for agent-based development
Format and what is included
- Discovery call: your industry, the team's daily work and the three targets of the day
- Content built on your own files, not a generic slide deck
- Full day (about 7 hours) hands-on, on site in Istanbul or online
- Team playbook with everything built during the day
- 30 days of Q&A support after the training
- Certificate of participation
Scopes beyond the standard package (more than 10 people, several locations, half-day or multi-day formats, integration work) are quoted separately.
The same standard package for every program.
How it works
15-minute call
We talk about the team, the work and whether this program fits. No commitment.
Discovery and preparation
A short session with the team lead. The day is built on your real tasks.
Training day
Everyone on their own laptop, on their own work, the whole day.
30 days of support
Questions that come up while actually using AI at work get answered.
Your trainer
Sefa Aydın
AI trainer and consultant. Years of producing design, advertising and software for the Türkiye projects of world-renowned luxury brands, founder of the agency Rebel Co. Group. I use these tools on real client work every day, and I teach what I use.
About me →Frequently asked questions
The team already uses Copilot. What does this training add?
There is a big difference between line completion and agent-based development. The program brings the team onto a working model that describes tasks, runs the agent within limits and reviews generated code by the rules. Using the tool and setting a team standard are not the same thing.
Is it safe to open our codebase to AI?
Yes, with a corporate account, a data processing agreement and a secret key rule. In the program, which code and data may enter a tool, how keys are protected and licensing questions are put into a written rule.
How much does AI training for software teams cost?
The standard package is a full day for up to 10 people at 49,900 TL plus VAT. The discovery call, content preparation based on the team's repository, a team guide and 30 days of Q&A support are included. Larger teams or multi-team organisations are scoped by proposal.
Which tool is better, Claude Code or Codex?
It depends on the job. The program shows the real differences between the four tools and the team tries them in its own repository. The current comparison is kept on the blog in the Antigravity, Cursor and Claude Code comparison and the coding tools comparison articles.
Will AI replace developers?
The developer who manages the code, not the one who types it, comes to the fore. The program teaches exactly this transition: task description, review, testing and security decisions stay with people, and recurring production is handed to the agent.
Can a product team that does not code take part?
Yes. The task description and acceptance criteria modules are especially valuable for product managers and non-technical stakeholders. If building applications without writing code is the goal, the vibe coding page is a separate resource.
Teams choosing this program also look at
Guides on this topic
Last updated: 14 September 2026
Request a proposal for your team
Tell me the team size and what you want to solve. I reply the same day.