Summary of "Бизнес-анализ с Chat GPT. Создаем беклог за 4 шага. ИИ заменит аналитиков?"

Overview

Business analyst Veronica demonstrates using ChatGPT to build a backlog in a four-step, structured workflow. The demo shows example prompts, expected outputs, and quality checks. Veronica frames AI as an assistant that can automate routine tasks and speed up work, while stressing that human oversight and core BA skills remain essential.

AI should be treated as an assistant to automate routine tasks and speed up work, not as a replacement for human judgment and BA skills.

4-step workflow (practical guide)

Preparation for interview
- Identify stakeholders and the information you need to collect (company website, meeting recordings, etc.).
- Upload context files to ChatGPT and ask it to generate a list of open-ended interview questions.
- Specify the interview purpose (for example: get a general idea vs. obtain deep understanding).
Validate and refine questions
- Review ChatGPT’s generated questions and edit as needed.
- Provide specific feedback when asking for a regeneration (treat prompts like instructions for an assistant—be explicit and iterative).
Convert interview answers into user stories
- Upload interview responses and ask ChatGPT to draft short user stories in a given format (title, short description, “who/what/why” / value-focused).
- Apply INVEST principles to each story:
  - Independent
  - Negotiable
  - Valuable
  - Estimable
  - Small
  - Testable
Structure backlog, identify MVP, and generate acceptance criteria
- Ask ChatGPT to create a User Story Map grouped by key user-journey actions to visualize scope, sequence, and the MVP.
- Request suggestions for stories or scenarios you might have missed (alternative flows, edge cases).
- Generate acceptance criteria in Gherkin format and ask that they follow SMART and other requirements-quality criteria:
  - SMART: Specific, Measurable, Achievable, Relevant, Time-bound
  - Additional checks inspired by requirements quality frameworks (e.g., BABOK-style quality checks)

Prompts, outputs, and tips

Always provide clear context and upload source files (website text, meeting recordings/transcripts).
Request open-ended questions for interviews; prefer feedback-driven regeneration rather than a generic “redo.”
Use ChatGPT to identify missing scenarios and to cross-check for gaps.
Convert AI outputs into visualizations or storytelling formats for stakeholder communication (e.g., story maps, diagrams).
Remember this speeds work and reduces omissions, but it still requires time and iterative refinement.

Quality control and human judgment

Never trust AI blindly: verify each output against domain knowledge and stakeholder context.
Perform a human assessment after every AI response to validate correctness, relevance, and adequacy.
Edit and adapt AI-generated artifacts to add authenticity, narrative, and context for stakeholders.

Skills AI won’t replace (per Veronica)

Emotional intelligence: handling stakeholder emotions and communication nuances.
Prompt engineering: the ability to craft clear, structured prompts to get quality AI output.
Critical evaluation / human judgment: interpreting and validating AI results with business context and experience.

Tools, frameworks, and formats mentioned

ChatGPT (GPT chat) — used in the demo
User Story Mapping
INVEST criteria for user stories
Gherkin format for acceptance criteria
SMART goals
Requirements quality criteria (reference to BABOK-style checks)

Practical takeaways

Start integrating AI into BA tasks: interview preparation, story generation, mapping, and acceptance criteria creation — but maintain rigorous review and domain validation.
Use AI to surface missed scenarios and speed routine work, freeing time for complex analysis and stakeholder engagement.
Invest in prompt engineering and core BA skills to remain effective and irreplaceable.