Authors : Milind Tailor and Aashima Gupta

Estimated Time : 5-7 mins

Manage it before it breaks your P&L

A year ago, AI tokens weren’t on your radar. Neither was your exposure.

Today – millions already spent. Zero visibility into what drove it. No governance model in place. No benchmarks. No contractual controls. Finance teams staring at invoices they cannot interrogate. Budgets exhausted months ahead of schedule.

Cast your mind back to 2017. Cloud spend was doing exactly this – growing faster than governance, bypassing procurement, and landing on finance desks as invoices nobody could explain. It took most organizations three to five years to build the visibility, the tooling, and the commercial discipline to manage it properly.

AI token spend is following the same arc. Except the timeline is compressed. The spend is more opaque. The billing is buried deeper. And the P&L exposure is arriving faster than anyone planned for.

You have been here before. The question is whether you act earlier this time.

Why This Category Is Different

Three structural facts make AI token spend uniquely difficult to govern.

No natural ceiling. Most cost categories have an anchor – a headcount, a license count, or a fixed contract. AI token spend has none of these. It scales with usage across teams and tools simultaneously, and nothing in the model naturally slows it down.
The cost is hidden by design. Spend sits inside your Azure or AWS bill as compute. It is bundled into SaaS license fees. It powers features inside tools your teams use every day and none of it surfaces as AI spend. It just looks like software working.
The market is not moving in your favor. AI pricing models are still evolving. Suppliers decide which models are used, how consumption is measured, what is included in subscription fees, and when additional usage charges apply. Commercial terms, reporting, and pricing structures vary widely across suppliers. This is not simply a visibility challenge – it is a commercial challenge. Without transparency, procurement cannot effectively benchmark suppliers, forecast costs or negotiate from a position of strength.

The Billing Has Three Layers, and Most Organizations Can Only See One

The way AI token costs flow through your business is not straightforward. There are three distinct layers, and most procurement and finance teams have visibility into only the first.

Layer 1 – Direct API contracts. The most visible. Your organization has a direct agreement with an AI provider, usage is tracked, and invoices are attributable. This is the layer most teams think of when they talk about AI spend. It is also the smallest part of the picture.
Layer 2 – Hyperscaler invoices. Token consumption embedded inside your Azure, AWS, or Google Cloud bill presented as compute or platform services. There is no obvious AI line item. No model names. No token breakdown. Just a number on a cloud invoice with little indication of what generated it or who consumed it.
Layer 3 – Embedded AI Within SaaS Platforms. The fastest-growing and often the least transparent layer. Enterprise software vendors increasingly consume AI on your behalf, either bundling the cost into license fees or passing it through as an additional consumption charge. In many cases, organizations are paying for AI capabilities they never explicitly purchased, at prices they never directly negotiated.

Most organizations are managing Layer 1 while Layers 2 and 3 accumulate unchecked. That is where the real exposure sits.

One line. £187,400. No breakdown, no owner, no prior approval on record. This is not an edge case – it is the standard invoice format for hyperscaler AI services today

What Procurement Does NOW

Five practical actions. This Year. Before AI spend becomes tomorrow’s cloud spend problem.

Action 1. Identify Where AI Spend Is Entering the Business

The first priority is understanding where AI spend exists.

Start by identifying your largest AI-related suppliers and invoices, then classify them into three commercial models:

Direct AI providers (OpenAI, Anthropic, Gemini)
Hyperscaler AI services (Azure AI, AWS Bedrock, Google Vertex AI)
Embedded AI within SaaS platforms (Microsoft Copilot, Salesforce, ServiceNow, Workday, etc.)

For most organizations, these three categories will represent the majority of AI-related spend and provide an immediate starting point for commercial oversight.

Outcome: Procurement gains visibility into where AI spend is concentrated and where commercial attention should be focused.

Action 2. Challenge Every AI Renewal for Commercial Transparency

AI is fundamentally changing the commercial model of enterprise software

Every renewal should now answer questions such as:

Is AI included in the license or charged separately?
What usage is bundled?
What triggers additional charges?
How will AI consumption appear on invoices?
What reporting and usage analytics are available?
What commercial controls exist to manage future growth?

Many vendors won’t proactively disclose this information. Procurement should ask for it.

Outcome: Greater transparency, fewer commercial surprises, and improved control over future AI costs.

Action 3. Assign Clear Commercial Ownership

One of the biggest risks isn’t overspending – it’s that nobody owns the spend. Every significant AI contract should have a clearly identified business owner responsible for:

monitoring usage
reviewing invoices
approving renewals
forecasting future consumption
working with procurement on commercial negotiations

Ownership creates accountability before governance becomes necessary.

Outcome: Every major AI contract has a clearly accountable business owner.

Action 4. Build a Common AI Spend Taxonomy

AI costs often appear under cloud, software, infrastructure, or departmental budgets.

Without a consistent taxonomy, procurement cannot benchmark spend, identify trends, or compare suppliers effectively.

A simple classification framework creates a common language across procurement, finance and technology. The primary lever differs by bucket but FinOps disciplines (tagging, showback, chargeback, waste identification) apply across all three.

Outcome: AI spend becomes measurable, comparable and easier to govern.

Action 5. Establish Executive Visibility

AI spend should become a standing agenda item not an annual surprise.

A quarterly review between Procurement, Finance and IT should cover:

largest AI suppliers
spend trends
upcoming renewals
commercial risks
areas requiring executive decisions

Regular executive visibility enables organizations to identify issues early, make informed commercial decisions and prevent AI spend from becoming another unmanaged technology cost.

Outcome: AI spend becomes an actively managed enterprise category rather than a reactive budget issue.

What Procurement Builds in the Mid-Term

The next 6-18 months. From visibility to commercial control.

By this stage, procurement understands where AI spend exists, who owns it, and how it is being consumed.

The next challenge is to ensure AI is procured, negotiated, and governed with the same commercial discipline as cloud infrastructure and enterprise software.

The four priorities could be:

Standardize AI Contractual Controls

The first generation of AI contracts focused on gaining access to AI capabilities. The next generation will focus on controlling commercial risk. Every significant AI agreement should begin to include standard commercial provisions.

AI usage reporting
Spend and consumption alerts
Price protection against future increases
Notification of pricing model changes
Audit and transparency rights
Consumption caps or budget controls where appropriate

Benchmark AI Commercial Models

One of procurement’s biggest challenges is that AI pricing is rarely comparable.

Some vendors charge:

token-based pricing
per-request pricing
AI credits
premium models
feature packs
bundled unlimited AI
fair-use policies

Comparing price alone is no longer enough.

Procurement should build commercial benchmarks that compare total cost of ownership, pricing models, bundled entitlements, reporting capabilities and vendor mark-ups.

Those benchmarks become the foundation for future negotiations.

3. Optimize AI Consumption

Reducing AI costs isn’t only about negotiating lower prices. It’s about reducing unnecessary consumption.

Procurement should begin asking vendors how they optimize AI usage through capabilities such as:

intelligent model routing
prompt optimization
response caching
duplicate request elimination
usage controls and budget limits

Two vendors may deliver the same business outcome, yet one may consume significantly fewer AI tokens through a more efficient architecture.

As AI adoption grows, procurement should evaluate vendors not only on what they charge for AI, but also on how efficiently they consume it. Efficient AI architecture is becoming a new commercial differentiator.

4. Establish Cross-Functional AI Governance

AI spend is no longer owned by one function.

Procurement, Finance, IT, Security, Legal, and business teams all have a role to play in ensuring AI is adopted responsibly and managed commercially.

Rather than creating another governance committee, organizations should establish a recurring commercial review that focuses on:

AI spend trends
new AI investments
major renewals
commercial risks
supplier performance
future demand forecasts

The objective is shared visibility, better commercial decisions, and stronger cross-functional alignment.

The Long Run – Direction of Travel

The next 2–3 years. From commercial control to strategic influence.

As AI becomes embedded in every business process, procurement’s role will naturally evolve. The organizations that lead this category won’t necessarily have the most advanced AI technology.

They’ll have procurement teams that built credibility early and became trusted commercial advisors to the business.

There are four long-term shifts.

Procurement Becomes the Commercial Advisor for AI

Procurement moves beyond negotiating contracts.

It becomes a strategic advisor, helping the business make better commercial decisions around:

commercial models
supplier selection
build versus buy decisions
enterprise licensing strategies
long-term AI investment planning

Commercial expertise becomes as important as technical expertise.

AI Becomes a Managed Enterprise Utility

Just as organizations manage cloud infrastructure and telecommunications, AI becomes another enterprise utility that requires continuous monitoring and optimization.

Success will no longer be measured by individual contracts, but by the organization’s ability to optimize AI consumption across hundreds of applications, thousands of users, and multiple suppliers.

Managing AI becomes an ongoing operational discipline, not a series of isolated procurement exercises.

Commercial Decisions Become Part of AI Architecture

Technology teams will decide how AI is implemented.

Procurement should help influence how AI is commercialized.

Questions such as:

Which commercial model scales best?
Should AI be bundled or metered?
Which vendors provide the greatest pricing transparency?
Which platforms minimize long-term operating costs?

Become part of enterprise architecture discussions, not just procurement negotiations.

Commercial decisions made during technology selection will increasingly influence AI costs for years to come.

AI Spend Becomes a Strategic Procurement Category

Cloud infrastructure created FinOps.

AI will create its own commercial discipline.

Organizations will continuously benchmark suppliers, forecast AI demand, optimize consumption, negotiate commercial models, and govern AI economics with the same maturity that exists today for cloud spending.

The organizations that start building these capabilities now will have a significant commercial advantage over those reacting only after costs accelerate.

The Bottom Line

Every major technology shift creates a new spend category.

First comes rapid adoption. Then accelerating consumption. Visibility lags behind. Commercial models evolve faster than governance. Eventually, finance starts asking questions the organization struggles to answer.

Cloud followed this path. Organizations spent years building FinOps capabilities to bring visibility, accountability and commercial discipline to cloud infrastructure.

AI token spend is following the same trajectory but at a much faster pace.

The difference is that procurement has seen this movie before.

The organizations that manage AI spend most effectively won’t necessarily be those with the most ambitious AI strategies or the lowest token prices. They’ll be the ones that establish visibility early, negotiate

commercial transparency, optimize consumption, and build governance before AI becomes another unmanaged technology expense.

AI is rapidly becoming a permanent enterprise operating cost not a temporary innovation budget.

The opportunity for procurement is not to slow AI adoption. It is to ensure AI scales with commercial discipline, transparency and sustainable economics.

Those organizations that begin building these capabilities today will be far better positioned than those trying to regain control after AI costs have already accelerated.

“The question is no longer whether AI will become a major enterprise spend category. It already has. The question is whether procurement will shape how that spend is governed or inherit it after the costs have already escalated.”

About the Authors

Milind Tailor

Global Procurement & Supply Chain Executive

Milind Tailor is a senior procurement and supply chain executive with extensive experience leading global procurement transformation, strategic sourcing and enterprise technology spend across multinational organizations. Throughout his career, he has helped organizations optimize complex procurement operations, drive commercial excellence and deliver large-scale business transformation.

Aashima Gupta

Founder & CEO, Sourcing Acumen

Aashima Gupta is the Founder & CEO of Sourcing Acumen, an AI-powered Source-to-Contract platform. She has deep experience in procurement transformation, strategic sourcing, and enterprise procurement technology, helping organizations modernize sourcing processes and adopt digital procurement solutions.