Case study analysis with AI: let it summarise, keep the judgement

1What AI can summarise

Where your course allows it, a tool can help with the mechanical part of a long case: listing the stakeholders, building a timeline of events, pulling the figures into one place, or reminding you of course theories that might apply. These save time without making decisions for you.

2What it cannot judge

USC Libraries' guide to case analysis explains that cases rarely have one correct answer, and that a case scenario may deliberately include misleading or irrelevant information to test whether students can sort through it. That sorting is the skill being assessed.

A tool summarising the case has no way to know which details are planted. It also tends to recommend balanced, general options rather than a specific, measurable plan. The marks go to your diagnosis of the core problem, your comparison of options and your recommendation, as set out in how to write a case study analysis.

3Checking numbers and facts

Check every figure a tool gives you against the case itself. Errors in numbers are a known weakness: in Walters and Wilder's 2023 study of ChatGPT citations, the most common errors in real references were in years, volumes and page numbers. In a case analysis, a wrong percentage or date can undermine your whole argument. If a tool adds facts that are not in the case, leave them out: many case assignments ask you to work only from the case and course readings.

4An example of a planted detail

Imagine a case, invented for this page, about a café losing customers. The case mentions that a competitor opened nearby, that the café raised prices, and, in a single line, that its opening time moved from 7am to 9am. A summary tool is likely to lead with the competitor, because that is the most common explanation for falling sales. A careful reader notices that the timing of the drop matches the later opening time, and that the lost customers were commuters.

That is the kind of judgement a case analysis rewards: spotting which fact explains the problem, and arguing for it with evidence from the case. No summary can make that call for you, because it depends on weighing details against each other. If you use a tool, use it after you have read the case, and test its summary against your own reading rather than the other way round.

5A workflow

1. Read the case yourself twice before using any tool.
2. If allowed, use a tool to build a list of stakeholders and a timeline, then check both against the case.
3. Write your own statement of the core problem, with evidence from the case.
4. Compare two or three options yourself, using course theories.
5. Write the recommendation and implementation plan yourself.
6. Record any tool use and acknowledge it as your course requires.

UCL's guidance, for example, lists producing summaries and analysing content among its examples for assessments where AI is an integral part, so check which category yours is in. For how these tools produce text, see how AI writing tools work; for longer coursework, term papers with AI; and for the plagiarism question, is using AI plagiarism?

More on this in AI and writing.

6FAQ