AI for IT Teams — Advanced

Full-day practical training for IT professionals already using AI who want to embed it systematically across the whole SDLC.

• Length: 1 day (8 hours)
• Price: from 18 000 CZK / participant (ex-VAT) — always quoted individually per group
• Format: on-site, custom-built for your team, on demand
• Max participants: 15
• Who it's for: developers, analysts, architects, tech leads — IT professionals who already use AI and want to go further
• Prerequisites: basic experience with AI assistants (Copilot, Claude, Cursor) or having completed "AI for IT Teams — Fundamentals"

• Work with AI systematically across the SDLC — from requirements analysis through implementation to output validation
• Build complex outputs iteratively — breaking a feature into epics, user stories, acceptance criteria, API specs and diagrams with AI assistance
• Critically evaluate AI outputs — spot where AI sweeps contradictions under the rug, skips edge cases, or proposes unrealistic solutions
• Set up AI workflow for the team — context files, rules, shared configuration, governance basics

• Working with raw, real materials — emails, meeting notes, conflicting requirements from different stakeholders
• AI as an analysis assistant: extracting requirements, identifying conflicts, structuring into a BRD
• Decision point: AI typically ignores requirement conflicts — participants must spot them and decide how to handle them

• Turning the BRD into an epic and breaking it into user stories
• Granularity judgement: Copilot typically proposes stories too coarse or too fine
• Rule: "If a story can't be described by one acceptance scenario, it's too big"
• Acceptance criteria and test scenarios (Gherkin)

• Generating Mermaid diagrams (sequence, ER) from user stories
• Decision point: AI misses error handling and boundary systems — participants add them
• Drafting OpenAPI spec: endpoint naming, HTTP methods, error states
• AI proposes the happy path; participants add edge cases

• Comparing AI outputs to a Definition of Done template
• AI validation prompt: "List ONLY missing items. DO NOT produce a fixed document."
• Participants fix it themselves — AI provides feedback, not finished output

• Context files (AGENTS.md / CLAUDE.md) — how one file improves every AI output
• Copilot Rules and team-wide configuration
• Commands and skills — repeatable workflows as templates
• MCP — connecting AI to your systems (intro, not a deep dive)

• What can go into AI and what can't (source code OK, production data NO, personal data NO)
• Tool consolidation — not fragmentation
• AI playbook: a one-page rulebook, not a 50-page document
• Measuring impact: how to tell AI is helping (and how to report it)
• Discussion and Q&A

• Project day. Participants spend the whole day on one cohesive project — from raw notes to a finished spec with diagrams and an API design.
• Deliberately complicated inputs. Materials include conflicting requirements, missing information and hidden dependencies — exactly like real projects.
• AI as a draft, not a finished solution. Every step has a decision point where the participant must critically evaluate and edit the AI output.
• Validation at the end. Participants verify the quality of their outputs against a DoD template — and fix what AI missed.

• Not a rerun of the basics. We assume participants already use AI — we don't start from zero.
• Not a coding session. We focus on the analysis and planning phase — implementation is the topic for the end-to-end training.
• Not a governance workshop. Governance is covered practically and briefly, not as a standalone topic.

1. Before the training we go through your stack, tools, processes, and current level of AI adoption.
2. Materials are adapted to your domain context (fintech, e-commerce, enterprise…).
3. The DoD template is built around your standards and processes.