1. The Quick Answer
Short version: ChatGPT can produce a case study that looks finished, with tidy headings and confident prose. Against a marker's rubric it usually falls short. It lists a framework's parts instead of applying them, invents or misdates evidence, and gives recommendations that don't follow from its own analysis.
Treat it as a brainstorming tool, not an author. Markers reward judgement and evidence, not formatting, so an unedited AI draft often lands at a low pass or worse.
2. What Markers Actually Look For
Distinction-level work shows critical evaluation at every step:
- Problem definition: one or two sentences naming the core issue, not a company summary.
- Framework selection: a stated reason why SWOT, PESTLE or Porter fits this problem.
- Evidence-based analysis: case facts plus peer-reviewed or industry sources.
- Options and trade-offs: two or three alternatives, each with risks.
- Recommendation: one option, justified against stated criteria.
- Limitations: missing data, assumptions and implementation risks.
Why AI struggles with the thinking inside each step
- It pattern-matches a typical answer instead of reading your specific case.
- Figures, dates and quotes may be outdated, misattributed or invented.
- It can produce references that look real but don't exist.
- It hedges ("it depends") where markers want a committed position.
3. SWOT and PESTLE: Lists Without Judgement
SWOT
- Mixes internal and external factors, such as a market trend under Strengths.
- Gives ten bullets per quadrant with no ranking.
- Uses vague wording like "strong brand" with no data.
- Stops at description, with no TOWS matrix or strategic option.
Fix: keep three to five evidence-backed points per quadrant, rank them, and turn the top findings into strategic options.
PESTLE
- Pads in irrelevant factors just to fill every letter.
- Writes in a generic "global" voice instead of the case's actual market.
- Uses stale economic and regulatory data, which matters in areas like AI regulation, ESG reporting and trade policy.
- Skips impact and likelihood ratings.
Fix: keep only factors the case supports, name the market, add current data from annual reports or regulators, and rate each factor's impact.
4. Porter's Five Forces and VRIO
Porter's Five Forces
- Assesses the wrong market boundary, such as "technology" instead of a specific segment.
- Rates every force "moderate" to avoid committing to a view.
- Mixes up buyer and supplier power.
- Ends with no "so what" about profitability or strategy.
Fix: define the industry precisely, rate each force high, medium or low with evidence, and finish by stating what the pattern means for margins and strategy.
VRIO and Resource-Based View
- Labels almost every resource valuable, rare, inimitable and organised, which defeats the test.
- Confuses resources with capabilities.
- Never explains why something is hard to imitate (causal ambiguity, path dependence, legal protection).
Fix: run each resource through all four questions in order and say clearly where it fails.
5. BCG, Ansoff, 7S, Kotter and More
- BCG matrix: labels "Stars" and "Cash Cows" from assumption, not from market share and growth figures. Rarely discusses the matrix's limits.
- Ansoff: casually recommends diversification without discussing capability gaps or capital.
- McKinsey 7S: describes each element alone, so a strategy-structure mismatch goes unnoticed.
- Kotter and Lewin: summarises the steps like a textbook instead of showing which step the organisation failed.
- Balanced Scorecard: lists generic metrics that don't connect through cause and effect.
- 4Ps / 7Ps: describes an "average" brand, with pricing that ignores cost, competitors and willingness to pay.
Fix: use real or case-supplied figures, discuss limitations, and compare options on risk, resources and expected return.
6. Recommendations, References and Integrity
AI recommendations are written to sound balanced, not to be defensible. Watch for generic advice ("invest in innovation"), no cost, timeline or owner, no risk assessment, and a conclusion that merely repeats the introduction.
A strong recommendation names the option, ties it to the framework findings, cites evidence, outlines implementation and acknowledges risks.
Reference like a pro
AI can generate citations that look real but don't match any source, mix Harvard and APA in one list, and present old statistics as current. Verify every source yourself against the case text, annual reports or peer-reviewed literature.
Submitting AI-written text as your own may breach your university's policy, and detectors can also flag honest work. Check your institution's rules, and use outside help only as a study example.
| Element | Typical AI draft | Strong case study |
|---|
| Problem | Company summary | One precise, testable problem |
| Frameworks | Lists every heading | Applies relevant parts, with reasons |
| Evidence | Generic or unverified | Case facts plus current sources |
| Evaluation | Descriptive | Weighs options and trade-offs |
| Recommendation | Safe and vague | Specific, costed and risk-aware |
| Referencing | Mixed or inaccurate | Consistent and verifiable |
7. Final Checklist: Use AI Responsibly
Use AI for brainstorming and structure only, never as the source of analysis or evidence.
- Read the case twice and highlight facts, numbers and quotes yourself.
- Write your own two-sentence problem statement.
- Choose the framework and note why it fits.
- Ask AI for prompts only, such as questions about supplier power.
- Verify every fact against the case, reports or academic sources.
- Draft your own analysis, then ask AI to challenge it.
- Build the recommendation with costs, risks and success measures.
- Proofread and reference carefully.
Self-check before you submit
- Does every framework end with a "so what"?
- Is every statistic traceable to a source?
- Does the recommendation clearly follow from the analysis?
- Are the limitations stated honestly?
Reference papers from any service are for learning only. Write your own final submission.
