Real company cases
11 publicly documented uses of AI by real companies. Each page states only what the published sources say, then shows a suggested workflow written for this guide (not the company's confirmed internal system) and a small project to rebuild the idea yourself.
-
Finance RAG / knowledge assistant Intermediate
Morgan Stanley
Advisors needed fast, trustworthy answers from a very large body of internal research and procedures.
Rebuild it: Build a permission-aware policy assistant over 20 synthetic procedure documents. Effort 30–40 h.
-
Commerce Agents / customer support Intermediate
Klarna
High volumes of routine payment, refund and return questions across many markets and languages.
Rebuild it: Build a support assistant with read-only order lookup, a refund tool that requires approval, and human handoff. Effort 30–40 h.
-
Hiring Fine-tuning Advanced
Indeed
Job recommendations without a clear reason were less persuasive and harder to trust.
Rebuild it: Compare a prompted model with a LoRA-tuned small model for job-match explanations on synthetic profiles. Effort 40–60 h.
-
Education LLM product features Intermediate
Duolingo
Learners want explanations of their mistakes and realistic conversation practice.
Rebuild it: Build a language-practice tutor with three scripted scenarios and a teacher review queue. Effort 25–35 h.
-
Agriculture Computer vision Advanced
John Deere
Broadcast spraying applies herbicide to the whole field even where there are no weeds.
Rebuild it: Train a weed detector on a public agricultural image dataset and simulate nozzle decisions at different sensitivity thresholds. Effort 40–60 h.
-
Infrastructure Predictive ML Advanced
Google DeepMind
Cooling systems are complex and interacting; fixed rules leave energy savings unrealised.
Rebuild it: Build a cooling-energy predictor on synthetic sensor data with a safety-constrained recommendation simulator. Effort 30–45 h.
-
Customer support Agents / customer support Intermediate
Intercom
Support teams receive many questions already answered in help articles.
Rebuild it: Build a help-centre answer bot over 30 synthetic articles with escalation and a resolution-rate definition. Effort 30–40 h.
-
Life sciences Enterprise assistants / data analysis Intermediate
Moderna
Specialists across functions spend time on analysis, summarisation and document drafting.
Rebuild it: Build an evidence-review assistant over public scientific abstracts that extracts claims and cites sentences. Effort 25–35 h.
-
Retail RAG / knowledge assistant Intermediate
Lowe's
Associates must answer questions about thousands of products and projects, including outside their own department.
Rebuild it: Build a store-associate assistant over a synthetic catalog of 200 products with stock filters. Effort 25–35 h.
-
Developer tools Code assistants Intermediate
GitHub
Developers spend time on repetitive code; measuring productivity is difficult.
Rebuild it: Build a repository Q&A and test-suggestion assistant for a small sample .NET solution. Effort 30–40 h.
-
Manufacturing ML platform / computer vision Advanced
Toyota
Factory floors have many local inspection and monitoring tasks but few ML specialists.
Rebuild it: Build a mini self-service image-classification pipeline: upload labelled images, train, report, export ONNX. Effort 40–60 h.
No case studies match these filters.