Deep|LLM: Enterprise AI Adoption Survey (Vol. 3) – Diverging Spending Outlooks, with Production Use Cases Scaling Alongside More Granular Cost Controls Spending divergence: Enterprise AI budgets remain positive but uneven, with high-growth respondents expecting roughly 40% to 60% spend growth over the next six months as production apps scale. Large European logistics company A raised its FY26 pure-AI budget from about $600K to $650K to about $1M, while Large Asian gaming company A expects paid-model spending to fall from more than $14.3M a month to about $11.4M to $12.9M by year-end. Production use cases: Customer service, voice, recruiting and workflow agents are moving into production at selected companies. Large Australian pharmaceutical distributor A says its customer-service agent already handles about 30% of inquiries, while Mid-sized Middle Eastern insurtech company A targets a 20 to 30 times increase in voice AI volume over the next six months. Cost controls: Companies are moving from broad experimentation toward more granular controls, including model tiers, default model settings, caching, seat reclamation and quota approvals. Large UK legal services company A reduced monthly costs from about 1.8 times budget to about 1.2 times after changing default models and access permissions, while Large European logistics company A reports caching has cut model consumption by about 20% to 25%. ROI and open models: ROI is clearest where AI can be tied to revenue, avoided hiring or lower external service costs, while employee tools are still measured mainly by adoption, time saved and output quality. Open-source model use remains mostly testing-stage, though Large Asian gaming company A has local models at about 16% of token usage and Large Southeast Asian telecom operator A may migrate 30% to 40% of current closed-model API usage to open-source models during 2027 if quality meets requirements. Detailed Report