A data science article makes the case for applying scientific methodology to AI and ML platform decisions, pushing back against the "prompt in, slop out" culture where teams accept AI outputs without verification. The author argues that simply asking AI for comparisons between platforms produces shallow results that damage professional credibility.

The piece walks through a concrete example: determining whether to consolidate two ML platforms used for similar churn prediction pipelines. The approach involves transforming vague business questions into testable hypotheses with defined independent variables (the platform), control variables (data, algorithm, hyperparameters), and dependent variables (cost, accuracy). The author recommends running multiple tests at different times to separate signal from noise.

The author emphasizes that hands-on experimentation builds authority while AI-generated posts damage professional relationships. Rather than posting "This is where you should use Platform A over Platform B," the author suggests framing findings as experiments: "When we changed the platform to see how it affects cost while keeping the algorithm the same, our findings were..." The piece draws inspiration from a Croatian academic paper by Professor Mladen Šolić on scientific research methodology.