AI as an operating metric

Visa says its teams are consuming AI at production scale — nearly 1–2 trillion tokens a month as usage doubled in weeks — and the company is rewarding employees for building faster with AI, signalling that token consumption is becoming a measurable internal operating metric rather than a pilot experiment. That shift forces platform teams to treat model access, cost allocation and governance as part of service management, not just developer tooling. (businessinsider.com)

Visa says it used 1.9 trillion artificial intelligence tokens in March 2026, up from about 1 trillion in February, which means its internal use nearly doubled in a few weeks. The company told Business Insider it is now rewarding teams that can show artificial intelligence helped them finish work faster. (businessinsider.com) A token is a tiny unit of text an artificial intelligence model reads or writes, so counting tokens is like counting gallons at a fuel pump instead of just asking whether the car moved. Deloitte said in January 2026 that tokens are becoming a new unit of value for companies because spending rises and falls with usage, not with a fixed software license. (deloitte.com) That changes how a big company measures progress. Visa technology president Rajat Taneja told Business Insider the company is watching “volume of impact,” which shifts the scorecard from “did people try the tool” to “how much work moved through the tool.” (businessinsider.com) Visa is not a small software shop running a side experiment. In fiscal 2025, it processed 257.5 billion transactions on its networks, handled $16.7 trillion in total volume, and operated in more than 200 countries and territories, so any internal metric it promotes tends to get wired into real operating systems fast. (annualreport.visa.com) (investor.visa.com) The company has already been building around artificial intelligence for years. Forbes reported Rajat Taneja oversaw about $3.3 billion of investment in artificial intelligence and data infrastructure, and Visa has used hundreds of models, including about 100 working in real time, for fraud and risk work. (forbes.com) Now the newer generative tools are spreading inside the company too. Fortune reported in October 2025 that Visa had built more than 100 internal artificial intelligence applications and had more than 2,500 engineers working specifically on artificial intelligence. (fortune.com) Once usage is measured in trillions of tokens, model access stops looking like a nice developer perk and starts looking like electricity, cloud storage, or network bandwidth. Deloitte’s January 2026 guidance says companies need forecasting, cost controls, and finance operations discipline because token-based costs are variable and can spike unpredictably. (deloitte.com) That also pulls governance into the middle of day-to-day operations. Deloitte’s 2026 enterprise survey says companies need governance before scale and a “living” technology and data infrastructure, because wider worker access creates a gap between experimentation and durable enterprise change. (replyfabric.ai) Visa’s public product moves line up with that internal push. On April 2, 2026, Visa announced Intelligent Commerce Connect to help businesses plug into artificial intelligence-powered shopping, which means the same company counting tokens inside its walls is also building rails for artificial intelligence agents outside them. (investor.visa.com) The signal here is not that Visa likes chatbots. The signal is that a payments company with $40 billion in 2025 revenue is starting to treat artificial intelligence consumption like an operating metric that can be tracked, budgeted, rewarded, and managed across the business. (annualreport.visa.com) (businessinsider.com)

Get your own daily briefing

Scout delivers personalized news, insights, and conversations tailored to your role and industry.

Download on the App Store

Shared from Scout - Be the smartest in the room.