**The most counterintuitive finding is that spending more on AI buys nothing measurable, while using it more does.** I computed pairwise correlations in one query (step 0). I'm reading the column abbreviations as follows: `c_*` is Monthly_AI_Cost, `h_*` is AI_Usage_Hours_Per_Day, `p` is Productivity_Score, `s` is Satisfaction_Score, `ts` is Time_Saved_Hours_Per_Week and `acc` is Accuracy_Rating. - **Cost is unrelated to outcomes.** Monthly AI cost has a correlation of about 0 with every outcome: - productivity: −0.007 - satisfaction: 0.019 - accuracy rating: 0.024 - **Usage hours track outcomes strongly.** Daily AI usage hours correlate with: - time saved: 0.76 - productivity: 0.64 - satisfaction: 0.40 - **The outcomes move together.** Time saved and productivity correlate at 0.76, and productivity and satisfaction at 0.65. Accuracy rating and satisfaction correlate at 0.53. - **Income and age don't explain it.** Income vs. cost is −0.05, and age vs. productivity is −0.06. Higher earners don't spend more, and older users aren't less productive. Together these suggest that a paid subscription tier isn't what drives productivity, time saved or satisfaction here. Engagement is what matters. For the "is a monthly AI subscription justified?" decision, the data gives no support for paying more. **Caveats** - This is observational survey data, so the usage–productivity link could run either way, or reflect self-reporting. - Cost has only 273 non-null values out of 300 rows, so each correlation uses the rows where both fields are present. - Several of these fields are stored as text, so they were cast before use. - I computed linear correlations only. A threshold effect, such as free vs. paid tiers, could be hidden by a near-zero Pearson value. I haven't checked that.