AI in Business ROI Barometer: Data from 200 deployments (2024-2025)

This dataset offers a detailed analysis of **200 Artificial Intelligence deployments** (Generative and Classical) carried out in SMEs, midcaps and large companies. It documents the actual cost, implementation time and **Return on Investment (ROI)** observed in the field. **Collection context** Unlike reporting studies based on opinion polls, this dataset is based on **operational audits** of projects in production. It aims to provide decision-makers and researchers with reliable metrics to get out of the "Hype" and assess the real profitability of AI. <https://www.denisatlan.fr/barometer-ia-pme> **Key dataset figures (V1.0.0):** * **Volume:** 200 projects audited. * **Period:** 2024-2025. * **Mediated ROI observed:** 159% (with a high variance depending on the use case). ***Failure rate:** 17.5% of projects did not reach production or a positive ROI. **Content of data:** The CSV file includes 15 key variables, including: * "Industry": Sector of activity (Tech, Retail, Finance, Manufacturing...). * "Use _Case _Category": Type of application (Customer Support, Code Generation, Marketing...). * "Implementation _Time _Weeks": Actual duration of deployment. * "Cost_Initial" & "Cost_Run_Monthly": Cost structure. * "ROI _Percentage": Profitability calculated over 12 months. * "Primary _Model": Model used (GPT-4, Claude 3, Llama 3, etc.). **Dictionary of variables:** * Project _id : Anonymised unique identifier (e.g. P001) * "sector": Sector of activity (Industry, Finance, Retail...) * "company _size": Enterprise size (SMEs, midcaps, large) * _m _eur income: Annual turnover in million € * `ai _use _case`: Use cases (e.g. Vision, NLP, Predictive) * "investment _eur": Total project budget (€) * "King _percent": Return on Investment (%) * _to _deployment days: Total duration until production (days) * Human _in _loop : 1 = Human in loop, 0 = 100% auto