The $1 trillion AI blind spot
Your cheatsheet to managing the fastest-growing, most under-managed spend category
A couple of years ago, AI spend amounted to a handful of $20 per month ChatGPT subscriptions and a few low- or no-cost trials. That era already looks like a distant memory.
Today, companies are paying based on how much their employees use these tools, breaking every forecasting tool CFOs rely on. This shift from predictable seat-based SaaS contracts to volatile, token-based pricing has created a blind spot for finance teams.
Ramp data shows businesses' monthly AI spend grew 4x from February 2025 to February 2026. Gartner projects global spend on AI services and software will total about $1 trillion in 2026.
Among businesses on Ramp that spend on AI, the median company dedicates nearly 15% of its total software budget to AI tools. Yet much of this spend res undermanaged or totally unmanaged as finance leaders lack visibility into it. As AI's adoption curve outpaces that of every previous technology and spending explodes, so do the risks of not managing it.
The companies that will thrive in this new era aren't just the ones that embrace AI but those that can see, control, and optimize this spend. Here, we lay out three steps to help you do just that.
AI spend is exploding
Token-based pricing has broken every traditional forecasting tool
4x
year-over-year increase in monthly AI spend (Feb 2025 to Feb 2026)Source: Ramp transaction data
80%
of companies miss AI spend forecasts by 25% or moreSource: Mavvrik survey
78%
of IT leaders saw unexpected AI-related chargesSource: Zylo survey
Monthly AI spend growth
Indexed to February 2025 baseline (100)
Source: Ramp transaction data
FIG_01 Monthly AI spend growth indexed to February 2025 baseline. Ramp data shows median business AI spend grew 4x from February 2025 to February 2026.
Read methodologyStep 1
Improve visibility
Gaining a clear understanding of your AI spend is difficult, but it's worth the effort. Before we get into how to fix this issue, here's why this is a challenge:
THE CHALLENGES
One-off projects create blind spots
Most AI tooling doesn't start with a procurement request. It starts with verbal sign-off from a manager for a small experiment. Engineers sign up for API keys on company cards. A sales leader expenses a team AI subscription after the fact. The scope grows quickly as employees find more use cases, and the gaps multiply.
Invoice-based billing, which Ramp data indicates accounts for the majority of AI spend, means you don't see the size of the expense until days or weeks later. And if AP and employee expenses live in different systems, it's even harder to get a sense of total AI spend. It's no surprise that 80% of companies miss AI spend forecasts by 25% or more, a survey from Mavvrik found.
The vendor landscape is fragmenting
Companies are increasingly working with multiple AI vendors, further obscuring visibility.
A year ago, the AI vendor landscape was essentially OpenAI plus a few niche players. Today, Ramp data shows year-over-year adoption growth of 150% or more for three distinct AI vendors: Lovable, xAI, and Together AI.
Companies are testing multiple models head to head, experimenting with domain-specific solutions, and switching between them as capabilities evolve. Your strategy for improving visibility needs to account for this fragmentation.
The AI cost paradox
Token costs are dropping ~10x per year, but is your usage growing faster?
Your AI usage growth (year over year)
2x
10x (break-even)
20x
Token costs
10x
cheaper
Your usage
more AI calls
Net spend change
costs decrease
Source: Andreessen Horowitz, OpenRouter
FIG_02 The AI cost paradox: falling token costs can't keep up with surging usage. Adjust the slider to see where your organization might land.
What to do
Audit existing vendors
Put out a call to action for employees to come forward with any subscriptions, with no threat of repercussions, and dig through statements and reimbursements. Build a list of every active AI vendor, who owns it, and what it costs annually to start immediately finding savings opportunities.
Consolidate spend in one place
Once you have a clearer view of what you're spending on, start moving those payments to a single spend management platform. Then, you can use reporting tools within the spend system to monitor and analyze AI spend.
Use cards for early warnings
Card transactions offer an additional layer of visibility, showing transaction details at the moment of purchase so finance teams can respond quickly. This is especially valuable as AI costs can jump substantially month to month. An added benefit: many corporate cards offer cash back on these increasingly large transactions.
Create an AI spend category
AI shouldn't be a subcategory of "software" anymore; it deserves its own line in your chart of accounts. This standardization will make it much easier to answer a basic, increasingly important question from your CFO: how much are we actually spending on AI?
How Ramp helps: Ramp manages all your spend — cards and expenses, AP, and procurement — in one place so you can monitor real-time spend patterns in a single dashboard.
Step 2
Strengthen controls without blocking teams
Once you have better visibility into these expenses, the next step is bringing costs under control and making them more predictable. AI presents a few new challenges here.
THE CHALLENGES
A volatile billing model
Most AI platforms have adopted a token-based model that's based on usage, a departure from the subscription model that dominated software pricing for decades. These costs are only expanding as AI projects grow more complex.
Billing looks different with different vendors. The top AI models like Anthropic, OpenAI, and Google charge based on usage of output (response) and input (prompt) tokens, with prices varying substantially based on the exact model used. This makes pricing much less predictable: Ramp data reveals AI costs spike 50% or more about one in four months for the biggest spenders, who tend to see the most dramatic swings.
"AI spend is the first major cost category where the biggest risk isn't just overspending but not knowing how much you're spending at all," says Ramp's Dave Wieseneck, a former finance leader at multiple startups. "When purchases start as $20 subscriptions and scale to six figures through usage-based billing, the traditional quarterly review cycle is already too late."
Other AI-enabled vendors have shifted pricing models with seat-based premiums (Microsoft), mandatory price increases to bundles (Google), prepaid credits (HubSpot), and flat fees per outcome on top of seat-based premiums (Zendesk). These changes seem to be driving up prices across the board: SaaStr, a community of software leaders, reported that SaaS prices rose 11.4% year over year in 2025.
AI SaaS pricing calculator*
Toggle vendors to see how costs could increase across your SaaS stack
Company size
2,500 employees
50
20,000
Per employee per year
$1,920$1,920
Total annual AI surcharge
$1.3M$1.3M
Sources: Microsoft Copilot, Google Workspace AI, Salesforce Agentforce, Slack AI, Canva TeamsZendesk
FIG_03 Potential added annual cost of widely used tools with AI capabilities. Adoption rates reflect that not all employees use every tool. Toggle vendors on/off and adjust company size to model your scenario.
*Based on vendor's listed or reported prices. Actual pricing may vary.
Cumbersome processes encourage "shadow AI"
The urgency around AI means employees have even less patience for long, complex procurement cycles. Yet usage-based pricing models increase the risk of unmanaged spend.
When procurement cycles drag on, teams sign up for free versions of software that auto-convert to paid or use leftover credits in their personal account at work after hitting usage limits. This presents not only cost concerns but heightened security risks (non-enterprise subscriptions mean your company's data could be used in model training).
Ramp data shows shadow AI is a growing problem: The volume of reimbursements for AI-related transactions tripled year over year, driven by twice as many companies processing AI reimbursements.
What to do
Create AI experimentation budgets
Rather than forcing AI purchases through the typical software procurement flow, create specific AI tooling budgets for the entire company to remove barriers. You can tailor these to various departments, as different budgets make sense for different teams.
Issue virtual cards with built-in limits
Virtual cards can provide merchant restrictions (only preapproved AI vendors) and spend caps that ease cost concerns. These limits enforce themselves, meaning finance teams can sleep well knowing big bills aren't quietly brewing. To avoid overly tight controls, get notified when teams near spend limits so you can proactively check in.
Set tiered approval thresholds
Not all AI spend is equal: A $50/month individual subscription doesn't need the same scrutiny as a $25,000/month enterprise agreement. So set approval thresholds based on spend tiers, with automatic approvals for anything below a certain amount.
How Ramp helps: Ramp offers unlimited virtual cards with granular controls and spend programs that let you set up and scale AI-specific budgets in minutes. For larger contracts, Ramp Procurement makes it easy to adjust approval flows based on spend tier.
Step 3
Optimize AI spend for a strategic advantage
Once you have better visibility and control, you can focus on the final step: optimizing this spend. Here's what you may be missing:
THE CHALLENGES
Measuring the full ROI of AI's ROI
Only 29% of executives say they can confidently measure the ROI of AI investments, IBM found. Anecdotal evidence and personal experience tells them AI ismaking a difference, but they're still struggling to quantify the financial returns.
Some may not have set clear KPIs and baselines before adding AI tools, which can make it harder to quantify gains later. Others may have a narrow view of ROI focused on time savings and the dollars those saved hours represent. They might not be considering other factors like output per employee, avoidance of new expenses, revenue impact, and customer satisfaction.
Insights help right-size and earn favorable terms
When you can see exactly how much you're spending with each AI vendor, how that spend has trended over time, and which departments are power users, it gives you powerful leverage in negotiations. Yet most lack these insights, as two-thirds of businesses use the vendors' own tools to track AI costs, and 57% track these costs in spreadsheets, CloudZero says.
The AI vendor market is still fiercely competitive, so data that can help you earn discounts based on committed spend or volume can meaningfully lower costs. These numbers can also help you right-size contracts with AI vendors and benchmark pricing with industry peers. The savings opportunity will only grow as your AI spend continues to scale up over time.
What to do
Build a clear plan to measure ROI
As early as possible, collaborate cross-functionally to determine what KPIs would help comprehensively capture ROI for different tools and teams. For example, it's not just that finance saves 20 hours every quarter pulling together reports; if marketing can now execute 50% more campaigns per quarter, increasing sales pipeline by 30%, that should be factored in.
Take a close look at utilization
Pull usage data and ROI details for your AI vendors and compare those against what you're paying for. Flag any duplicate or under-used subscriptions with suggested next steps, like moving to a cheaper tier or not renewing the contract.
Run AI spend reviews every four to six weeks
Track AI spend and ROI even more closely than other major spend categories, given the volatility around billing and usage (employees' preferred tools can change quickly). Run regular surveys and value reviews with key teams to understand how ROI changes over time. This can also help you catch unfriendly (and costly) contract terms.
How Ramp helps:Ramp's native reporting offers spending and usage insights by vendor, team, and individual so you can proactively right-size or renegotiate agreements.
The road ahead
AI spend will continue to climb quickly: Ramp projects the average enterprise AI contract will hit $1 million this year, up from less than $150,000 in 2024. Companies that continue to treat it as just another line item and manage it ad hoc will soon fall behind.
"A year from now, AI will be one of your top three software line items," Wieseneck says. "The companies building the infrastructure to manage it today will have a compounding advantage via better data, better vendor leverage, and better forecasting."
Those winners are the ones that will move quickly to uncover AI spend, set up intelligent and flexible strategies to control this spend, and regularly revisit usage data to unlock additional savings. The tools to run this playbook exist. The question is whether your finance team will start saving now or scramble to control spend later.
$10B+
saved by Ramp customers
27.5M+
hours saved
50,000
businesses trust Ramp
Methodology
Data in this paper is drawn from anonymized, aggregated transaction data across Ramp's customer base. "AI spend" includes transactions with generative AI providers. Spend figures and payment method analysis reflect the trailing 12 months ending February 28, 2026, unless otherwise noted.