Token Economics: “Moneyball” for AI in Supply Chain
Like the overlooked stats that changed baseball, token data reveals the hidden patterns of AI value in supply chain operations.
As genAI moves from pilots to the core of how work gets done, the unit of value is shifting. Instead of tracking software licenses, FTEs, or active agents, executives are starting to think in tokens – the atomic unit of AI work. Done right, token-level thinking can become a powerful tool for delivering AI ROI. Done badly, it’s a new way to burn budget with little to show for it.
Understanding how to govern AI initiatives with an eye on token accounting is key to avoiding pilot purgatory and winning the productivity game.
The Rise of the Token Economy
In traditional SaaS, leaders manage cost around seats and modules. In the AI era, the key variable is increasingly tokens – the chunks of data that models consume and produce when they’re asked to do work.
Over the last month, this has gone from technical detail to boardroom topic:
• At NVIDIA’s GTC conference, CEO Jensen Huang argued that enterprises should expect heavy token use if they want truly AI-enabled engineers, going so far as to suggest he’d be “deeply alarmed” if a $500,000 engineer wasn’t burning tokens worth at least half their salary on AI each year.
• Box CEO Aaron Levie recently warned that AI token consumption will quickly move beyond data science and engineering into legal, sales, and operations, forcing companies to learn how to budget for “workers running up AI token bills.”
For CSCOs, tokens are becoming the metered fuel of AI-enabled work. The question is not whether your organization will spend on them, but whether the gains in individual productivity add up to better workflow performance across supply and demand as a closed-loop system.
Tokens Reveal How Teams Use AI to Work Better Together
Zero100’s research on AI ROI shows that leading organizations already think beyond pure cost. They frame AI around three ideas:
• Orchestration and cycle time as the primary marker of value – Our analysis of nearly 300 automation implementations found that 46.1% of companies lead with time-based metrics of orchestration (processing time reduction, hours saved, throughput) rather than narrow labor cost savings. AI is increasingly judged on improving cycle times, not just eliminating headcount.
• Balanced productivity across both demand and supply operations. Top quartile AI performers see ~2x revenue growth and 1.6% higher gross margins than laggards. They invest in AI across both supply and demand sides of the closed loop, and they deliberately connect functions using what we call PowerThreads (like Plan x Move at Unilever and Use x Make at Intuitive Surgical).
• System-level returns over task-level savings. We found that technology is only ~20-25% of automation cost. 75-80% is process redesign, integration, and change management. When leaders think in systems, they justify bigger, longer-horizon investments with outsized impact instead of chasing dozens of small, disconnected automations.
Token economics amplifies all three of these. By measuring who is using which AI tools, it’s possible to precisely monitor how teams work with AI to speed up and improve outcomes of selected cross-functional workflows. Different token volumes, for instance, on inputs vs outputs or between text-centric tools vs image-centric tools, help leaders see where AI is doing the work and where humans-in-the-loop are intervening.
Moneyball for Supply Chain
I recently had a conversation with a CSCO whose company was monitoring token use at both group and employee levels. He worried that marketing was using AI more than operations, feeling behind on AI adoption. In fact, the token consumption data probably revealed loads of artwork creation rather than spreadsheet digestion. Or, as they say in the data-rich world of sports, the devil is in the details, since output tokens are 5-10x the cost of input tokens, while graphics can be 60x as token-hungry as text.
Crunching the numbers behind token use of a cross-functional team (aka, PowerThread) lets a leader see exactly how and by whom AI is being used to get better, faster, or stronger. The Moneyball analogy says it all – learn to use token data to build and coach your workflow teams.
After all, running up gaudy stats while losing games is no way to win a championship.