KPMG surveyed 2,145 senior leaders for its Q2 2026 Global AI Pulse and found a gap on the cost side of AI that finance teams should take seriously.
Only 35% of organizations have full visibility into their AI operating costs. Nearly half have delayed or scaled back AI agent deployments after costs began to outweigh the benefits.
Organizations with full cost visibility report established ROI at five times the rate of those without it, 15% versus 3%. KPMG is measuring the cost side, and is asking whether organizations can see what their own AI spend is and what it produces.
For companies that intend to monetize AI, the same visibility problem will appear on the revenue side, and it will be harder to live with there.
If 65% of organizations cannot fully see what AI costs them to operate, how many can see whether every unit of AI-driven usage a customer consumes is priced, billed, recognized, and reported correctly?
Traditional enterprise software has finance more predictable footing. You buy licenses, negotiated contracts, budgeted annually, and costs held reasonably steady for the year. AI makes that footing far less stable, on both sides. Costs move with usage, demand shifts daily, pricing varies by model, and value is delivered continuously.
Finance teams are being asked to manage commercial models their processes were never built to support, and the KPMG data suggests most of them already know it.
We Have Seen This Before
Every major technology shift ends up testing the back office.
The dot-com era exposed weaknesses in fulfillment, inventory, and operations. SaaS exposed billing and revenue recognition processes built around one-time sales.
In both cases, the companies that struggled had working technology and broken commercial plumbing.
AI applies the same test with less patience.
For companies selling AI through usage-based, outcome-based, or hybrid commercial models, pricing, consumption, customer value, and revenue recognition are all in motion at the same time, and the systems connecting them either keep up or they fall behind quietly.
AI Is Stress-Testing Revenue Architecture
Two questions separate the organizations handling this well from the ones improvising.
Can you trace a dollar of AI-driven usage from contract through billing to recognized revenue without manual intervention? When consumption swings materially in a month, does your forecast move with it, or does someone rebuild a spreadsheet?
In the environments we encounter, the answer is rarely a simple yes to both. Billing, ERP, CRM, and reporting systems were aligned once, at implementation, around a business model that has since changed.
Every new pricing experiment adds another manual bridge between systems, and every manual bridge is a place where revenue leaks or the numbers stop matching.
What Readiness Looks Like
The advantage goes to companies that can meter consumption accurately, price new models without launching a systems project, bill correctly at any volume, recognize revenue in line with how value is actually delivered, and give their CFO numbers that hold up in a board meeting. That is a revenue architecture question, and it is answerable before the pressure arrives.
A Better Question to Ask
Most organizations are asking how fast they can adopt AI. The more useful question is whether you would trust your own numbers once you did. We call that state revenue confidence: everything you have contracted is being billed, recognized, and reported correctly, even while pricing and consumption move daily. Architecture is how you get there. Confidence is what the CFO is actually buying.
At Synthesis Systems, we help organizations prepare their revenue architecture for changing business models, not just new technologies. If you are working through AI monetization, usage-based pricing, or how your Quote-to-Cash systems will hold up as your business evolves, contact us below or email Sales@synthesis-systems.com.