Key takeaways:
- AI costs are already forcing hard business decisions. Across 396 companies, almost two-thirds said AI costs materially changed a decision. Kathy’s read is that most companies aren’t catching this early enough. They dive into AI, let usage spread, then only look back at the true cost once it’s already piled up.
- Everyone tracks AI spend. Almost no one can forecast it. Sundeep’s take: AI is inherently more volatile than traditional compute, and most of the spend is happening bottoms-up with zero governance.
- Most companies are only seeing part of the bill. Only 53% include hosted model and API costs in their AI cost reporting. Only 52% include data platforms and only 42% include developer tools. Kathy calls this the “visibility cliff” and the human cost of validating AI output rarely makes the list at all.
- You can’t measure ROI without a baseline. Kathy and Larry both hit the same point: most companies never documented what a process cost before AI, so there’s no way to prove it was worth it after. Sundeep recommends you get brutally specific about cost per outcome and value per outcome, then compare that to the status quo. If the math doesn’t work, get rid of it.
- Agentic AI has a governance gap most companies haven’t caught up to. 98% are piloting or running agentic workflows, but only 36% report those costs. Kathy’s says every company needs something an “AI Control Center,” with finance holding a real seat at the table.
Best for: CTOs | CFOs | Leaders responsible for AI strategy and adoption, cloud and infrastructure costs, FinOps and technology financial management, data and analytics, AI governance, risk, and compliance, and measuring the business value of AI investments