Bauplan’s 2024 summer internship program paired interns with ChatGPT to prototype a branching data platform for retail sales predictions. The experiment tested whether AI could replace human research spikes.

The team used ChatGPT‑5.2 for strategy, Claude Code for development, and Roborev for code review. Results showed the AI completed ~80 % of work in 48 hours, but relied on shortcuts that risked inaccurate outcomes.

Interns contributed insights that the AI missed, reinforcing that hands‑on mentoring produces deeper understanding. The prototype delivered a functional web app and query simulation, yet the learning curve for nuanced prompts remained.

Ultimately, the trial confirms AI can accelerate prototyping but not replace the critical thinking and iterative feedback that interns provide. The project highlights the need for balanced human–AI collaboration in data engineering.