Prajkta Waditwar-Senior Technology Sourcing Manager at Box, focused on AI strategy & Procurement Innovation. The views expressed are my own.
Over the course of my career, I’ve watched enterprise technology reinvent itself more than once.
I remember when organizations debated whether to move away from on-premises infrastructure. Later, the conversation shifted to whether SaaS could replace traditional software. Then came cloud computing, forcing enterprises to rethink not just technology but how they budgeted, governed and optimized infrastructure. That shift eventually gave rise to Cloud FinOps.
Each transformation changed the role of technology. None fundamentally changed the role of procurement. Until now.
Over the past year, as I’ve spoken with technology leaders, procurement professionals, architects and business executives across the industry, I’ve noticed a recurring pattern. Regardless of industry or organization size, the conversations often arrive at the same question:
“Why is our AI spending growing so much faster than we expected?”
What’s interesting is that the answer is rarely a failed procurement strategy or an expensive new contract. Instead, it’s usually a reflection of something much simpler. More employees are using AI. Developers are integrating LLM APIs into more applications. Marketing teams are generating personalized content at scale. Legal departments are reviewing contracts with AI. HR is drafting job descriptions in minutes instead of hours. Customer service teams are deploying AI assistants to support thousands of customer interactions every day.
No single event causes spending to increase. Millions of everyday interactions do. That observation completely changed the way I think about enterprise AI.
What I see changing is the thing organizations are actually paying for. The software may provide the interface, but the cost increasingly follows how much AI capability the business uses and how broadly that usage expands.
It led me to a simple conclusion: Every AI prompt is a procurement decision.
From Buying Software To Buying Intelligence
Technology procurement has always evolved alongside technology itself. First, enterprises purchased hardware. Then software licenses. Software-as-a-Service transformed ownership into subscriptions. Cloud computing shifted infrastructure from capital investment to consumption, ultimately giving rise to FinOps, a discipline focused on governing cloud economics rather than simply reducing cloud costs. Generative AI represents the next transformation.
Generative AI adds a different commercial dimension. Usage can expand with every new workflow, API integration or employee use case, which makes the economics much more variable than a traditional software license.
Every prompt submitted to ChatGPT. Every API request made to a foundation model. Every AI-generated report. Every autonomous agent completing a workflow. Every code suggestion generated by an AI coding assistant. Each consumes tokens. Those tokens may appear to be technical billing units, but from an enterprise perspective, they represent something much more significant.
They represent business consumption.
Industry research suggests enterprises are experiencing this shift rapidly. Deloitte’s State of Generative AI in the Enterprise highlights a clear shift in enterprise AI adoption from pilots and proofs of concept toward scaled deployments. As organizations move along that journey, the research points to governance, risk and compliance and demonstrating business value as critical priorities for successful scaling. Similarly, McKinsey’s State of AI highlights that organizations generating the greatest value from AI are redesigning business processes and operating models—not simply deploying AI tools.
Those findings are consistent with what I have been seeing in the market. As AI moves into day-to-day operations, the harder questions are increasingly about governance, cost, accountability and whether the use case is delivering enough value to justify continued investment.
Why I Believe We Need Token-as-a-Service (TaaS)
Traditional procurement was built for predictable technology investments. Generative AI isn’t predictable. Every major technology shift has created a new management discipline. ERP required enterprise architecture. Cloud computing gave us FinOps.
I believe generative AI now requires a new procurement discipline. That’s why I’ve been exploring the concept of Token-as-a-Service (TaaS).
I see TaaS as a procurement lens for managing AI consumption. Tokens provide organizations with a measurable way to understand how AI services are consumed, but the procurement challenge goes beyond the unit price. Leaders also need to understand demand patterns, contractual commitments, supplier exposure and the business outcomes associated with that consumption.
Managing them requires organizations to continuously forecast demand, diversify suppliers, optimize consumption, govern commercial risk and connect AI investments to measurable business outcomes. The objective shifts from buying technology to managing the economics of intelligence. That is a fundamentally different procurement challenge.
AI Procurement Is Becoming Portfolio Management
One aspect of AI that doesn’t get enough attention is supplier strategy.
A few years ago, most organizations standardized on a small number of enterprise software vendors. With generative AI, that’s becoming increasingly difficult. Different foundation models excel at different tasks-reasoning, coding, summarization, multilingual communication or image generation. New models emerge rapidly, pricing changes frequently, and capabilities improve at an unprecedented pace. For procurement, this may require a portfolio approach. Different providers and models may make sense for different workloads, so sourcing decisions need to account for performance, cost, security, compliance, switching flexibility and concentration risk over time.
That requires a fundamentally different procurement mindset.
The Next Competitive Advantage
One of the biggest misconceptions about enterprise AI is that lower AI spend equals better governance. It doesn’t. Token reduction by itself is a poor measure of success. I would rather see procurement understand what the organization is getting in return for that consumption and whether higher usage is producing better productivity, customer experience or business outcomes.
As frontier models become increasingly accessible, competitive advantage won’t come from having access to AI. It will come from managing it effectively. The organizations that succeed will be those that can continuously balance cost, performance, risk and business value.
That’s why I believe it’s time to think beyond traditional software procurement and toward Token-as-a-Service (TaaS)—a framework for governing enterprise intelligence as a strategic, consumption-based resource.
For me, the more useful question is what the organization receives for the AI capacity it consumes. As usage grows, procurement should be able to connect that spend to an outcome the business can actually recognize and measure.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?