AI budget overruns are getting harder to catch because AI no longer behaves like fixed software spend. It moves through seats, tokens, APIs, cloud calls, and agents that can keep running long after the budget owner stops watching.
Bloomberg reported that Uber capped employee usage of agentic coding tools after maxing out its entire 2026 AI budget in just four months. Axios reported that one enterprise client spent half a billion dollars in a single month after failing to put usage limits on Claude licenses. These aren’t normal software overruns. They’re warnings about what happens when token-based and agent-driven AI costs scale faster than finance can see them.
Key Takeaways
- Most enterprises can’t connect AI spending to specific teams, tools, use cases, or outcomes, which makes AI budget overruns difficult to prevent.
- Usage-based AI costs, especially from agents and high-volume token consumption, can increase much faster than traditional software costs.
- Getting one accurate picture of total AI spend is the first step toward controlling it.
The Problem Nobody Expected
A year ago, most finance teams were focused on software seat licenses. The math was simple: count seats, multiply by price, and budget accordingly. AI broke that model.
Today, AI expenses come from several directions at once:
- Software as a Service (SaaS) seat licenses for tools like Copilot and ChatGPT Enterprise
- Pay-per-token Application Programming Interface (API) costs that increase with usage
- Cloud model calls billed through infrastructure
- Autonomous agents that can keep running overnight, over weekends, and around the clock
The Token Spend & Insights dashboard was built to consolidate that fragmented picture.
According to Zylo’s 2026 SaaS Management Index, 78% of information technology (IT) leaders have encountered unexpected charges tied to AI or consumption pricing. That number isn’t surprising when you understand the structure of the problem. The issue isn’t that teams are spending carelessly. It’s that visibility tools haven’t kept up with how AI is actually billed.
Where the Money Goes Missing
Larridin scan data across enterprise clients shows an average of 18% of AI spend is unattributed. That means nearly one dollar in five can’t be connected to a team, tool, use case, or business outcome.
Spend typically disappears in four places.
Agents Running Without Oversight
Autonomous agents don’t work banker’s hours. A single misconfigured or orphaned agent can burn through a team’s quarterly budget while everyone is asleep. This is especially risky when agents can trigger chained workflows, make repeated model calls, or keep working without a clear owner.
Shadow AI Subscriptions
Teams are solving real problems with AI tools that never went through procurement. Those costs land on personal credit cards, team expense accounts, or departmental software budgets with no central visibility.
In Larridin enterprise audits, we consistently find three to five times more AI tool usage than IT expects. The full picture of shadow AI is one of the most consistent surprises for leadership teams.
Embedded AI in Existing SaaS
Most major productivity suites have added AI features in the last 18 months. Some are included in existing contracts, but many are not. Finance teams often don’t have a clear view of which AI capabilities are included in tools they already pay for.
No Single Total
Licenses, tokens, API calls, cloud model usage, and agent activity are all measured differently and appear in different systems. Without a tool that reconciles them, answering “What are we spending on AI this month?” means pulling data from four or five places and hoping nothing is missing.
What Real Spend Visibility Includes
Getting control of AI spend doesn’t require limiting what teams can do. It requires knowing what’s happening.
A complete picture of enterprise AI spend connects to the broader AI measurement and optimization framework of utilization, proficiency, and value. Spend is the input. Outcomes are what justify it.
Strong AI spend visibility should show:
- Total spend by tool, team, role, and department, updated before the month closes
- A clear split between human-driven spend and agent-driven spend
- Attribution of every dollar to a use case, owner, and output metric
- Early warning signals when a tool, team, or agent is on track to exceed budget
What Finance and IT Leaders Should Do Now
- Establish a single total first. Before optimizing, get one number for AI costs this month across all sources.
- Assign every dollar to an owner. Unattributed spend is ungovernable. Every tool, agent, and seat needs a team and a use case.
- Treat agents as a separate budget line. Agent costs are more variable than seat licenses, so they need their own visibility, thresholds, and accountability rules.
- Build before-and-after baselines. Without a baseline, there is no return on investment (ROI). The CFO guide to AI monitoring walks through how to structure that measurement.
The window to get ahead of this is now. Bain’s 2026 Automation and AI Pathfinder Survey found that nearly 40% of companies that measured AI cost savings landed below 10%, despite targeting 11% to 20%. The difference between the companies that hit their targets and those that missed was rarely tool access alone. It was whether leaders could connect investment, usage, workflow reality, and outcomes.
Frequently Asked Questions
Why is AI spending so hard to track?
AI spending is hard to track because it comes from multiple sources that don’t share the same billing model or reporting system. Seat licenses, token consumption, cloud model API calls, and autonomous agent activity all show up in different places. Without a tool that consolidates them, finance teams are working from an incomplete and often understated picture.
What is an orphaned AI agent?
An orphaned AI agent is an autonomous AI system that’s actively running and consuming resources but has no current owner attached to it. These agents are often created during pilots or projects and left running after the project ends. They’re a common source of unplanned AI budget overruns.
How much of enterprise AI spend is typically unattributed?
Larridin scan data shows an average of 18% of enterprise AI spend can’t be attributed to a specific team, use case, or outcome. In some organizations, particularly those that moved quickly on AI adoption without establishing measurement first, that number is significantly higher.
What is the fastest way to reduce AI budget overruns?
Start with a discovery scan to surface all AI tools in use, including shadow AI. Establish a single total across all spend sources. Then attach every dollar to an owner and a use case. That sequence creates the foundation for budget controls, and it usually surfaces enough orphaned spend to fund the next round of strategic investment.
Get Visibility Before the Next Budget Review
You can’t govern what you can’t see. Larridin gives finance and IT teams a single view of AI spend across every tool, team, agent, and workflow, so you can stop reacting to overruns and start making decisions based on evidence.
Book a discovery call to see what your AI spend actually looks like.