Publisher : Laurentiu Raducu
Price : $14
Course Language : English
Description
Business analysis is mostly reading, listening and writing: interviews, workshop notes, requirements, process maps and status updates. AI is very good at a lot of that work, and dangerously bad at some of it. This course shows you which is which, on one realistic project, with real results.
You follow a freight company's Delivery Exceptions Portal from the first brief to a prioritised backlog. You get the brief, three stakeholder interviews, a workshop transcript, a claims policy, a vendor offer, a draft requirements document and a quarter of ticket data. And the project has traps on purpose: two stakeholders who want opposite things, phrases that sound good and mean nothing, open tickets that distort a careless average, and a document with a hidden instruction for AI.
What you'll do with AI:
Prompting that works: context, task, format and limits, shown on the same interview summarised twice.
Elicitation: interview guides, summaries with exact quotes (checked word for word), conflicts between stakeholders, questions for a vague brief, and minutes with real owners and dates.
Requirements: functional requirements with sources, user stories with Given/When/Then acceptance criteria, a strict ambiguity review, non-functional requirements and a traceability table.
Processes and models: as-is and to-be processes as Mermaid diagrams, gap analysis, decision tables from a policy, and a stakeholder map.
Data analysis: AI that runs code instead of guessing, SQL explained in plain English, and every number recomputed independently.
Documents: a requirements document section, one update for two audiences, MoSCoW prioritisation and a backlog ready to import.
Risks: testing for invented facts, redacting confidential data before prompting, hidden instructions in documents, and inconsistent answers.
Your way of working: a prompt library, a nine-point review checklist, and what stays your job.
Real prompts, real answers. Every prompt was sent to Claude and every answer is shown as it came back, with the time and the cost: 32 recorded runs, for less than one dollar in total. When a result surprised me, you see that too, such as the plausible additions that slip into otherwise excellent answers, and the data error Claude found that I hadn't planted.
No programming needed. All case study files are included as a download, so you can rerun every prompt yourself.
Recorded with Claude Sonnet, October 2026. About 1 hour 40 minutes in 35 lectures of two to four minutes, each with a short recap.