Overview
- This week at Gartner’s Data & Analytics Summit analysts urged IT leaders to stop treating models as the sole priority and instead invest in governance, change management, data quality and workforce reskilling to turn AI experiments into business value.
- Consumption-based pricing tied to tokens and GPU hours, plus agentic workflows that multiply model calls, are producing surprise invoices and budget overruns and have prompted warnings that AI coding costs could outpace average developer salaries by 2028.
- Rapid prototyping with AI is accelerating feature delivery while creating architecture debt and reliability gaps, with developers often shipping AI-generated infrastructure code with minimal review and nearly all organizations reporting at least one AI-caused infra incident.
- Practical responses are emerging: companies are standing up AI FinOps teams, embedding policy-as-code into pipelines, exploring private inference and using platform engineering to make governed workflows the faster path to production.
- Analysts say model performance is converging so lasting advantage now depends on integrating high-quality data, clear business context, strong engineering practices and AI literacy; without those investments many pilots will be abandoned and staff will face burnout or reskilling demands.