Rethinking Corporate Finance: How Esker Is Embedding AI Token Consumption Into Employee Expense Planning

As artificial intelligence shifts from an experimental novelty to a core operational driver, corporate finance departments are grappling with an unprecedented budgeting crisis: unpredictable, skyrocketing costs. For finance software provider Esker, the financial reality of widespread AI adoption hit hard this year, with expenditures soaring to roughly four times initial projections.

In response, the company is pioneering an innovative accounting framework, treating AI token consumption much like salary, benefits, and payroll taxes—essentially bundling artificial intelligence into the total cost of an employee. This novel approach highlights a broader, industry-wide struggle: how to harness the immense potential of generative AI without losing control of runaway usage-based expenses.

Main Facts

The core challenge facing modern enterprises is the unpredictable nature of usage-based pricing models deployed by AI providers. As models become more sophisticated, intuitive, and embedded in daily workflows, corporate reliance on them deepens.

• The Budget Overrun: Esker’s AI costs surged to roughly four times over budget this year as the company aggressively ramped up its integration of the technology.
• The Token Problem: AI models process data in "tokens"—the basic units of text or code. As companies increase their usage, they routinely exceed their pre-set token caps, causing monthly expenses to spike unpredictably.
• The Employee-Centric Solution: To regain control, Esker’s Chief Financial Officer, Scott McDermott, has integrated AI token consumption directly into employee expense planning. By calculating a run-rate AI cost per employee across different departments, the finance team can track consumption patterns and forecast future budgets with greater accuracy.
• Departmental Disparities: AI consumption is far from uniform across the enterprise. According to McDermott, research and development (R&D) and finance teams are the heaviest consumers, frequently driving up overages compared to other departments.
• The ROI Dilemma: While tracking the cost side of the equation is becoming more systematic, linking that expenditure directly to tangible business outcomes—particularly revenue growth—remains a persistent hurdle for leadership.

Chronology of the AI Cost Crisis

The path toward Esker’s current budgeting strategy is part of a rapid, widespread corporate evolution in artificial intelligence adoption over the last several years.

Early Adoption and Free-Form Experimentation (2023–2024)

During the initial wave of generative AI excitement, most organizations approached artificial intelligence as an auxiliary tool. Software deployments were treated like standard IT subscriptions—flat-rate or lightly metered expenses managed primarily by tech departments rather than corporate finance. Budgets were small, usage was experimental, and token overages were negligible.

The Usage-Based Pivot and Scaling Realities (Late 2024–Early 2025)

As AI models demonstrated undeniable productivity gains, corporate reliance accelerated rapidly. Recognizing the shifting market, AI model and infrastructure providers universally transitioned away from flat subscription tiers toward dynamic, usage-based, per-token pricing structures. Companies began integrating AI deep into daily workflows, software suites, and automated tasks, causing consumption to skyrocket.

Leadership Transition and the Budgetary Awakening (January 2025)

In January 2025, following more than two decades in corporate finance, Scott McDermott stepped into the role of CFO at Esker. Almost immediately, he confronted a stark financial reality: the company’s AI token consumption was consistently blowing past its caps, with expenditures running roughly four times over budget.

Developing the Employee-Expense Framework (Mid-to-Late 2025)

Realizing that traditional IT or software budgeting models were inadequate for managing dynamic AI costs, McDermott and his team began formulating a new accounting strategy. By mid-2025, Esker started pulling actual usage data from its AI vendors to map consumption directly to individual employee roles and functional departments.

Industry Recognition and the 2026 Reality

By 2026, the issue had become an industry-wide phenomenon. A September 2026 company survey revealed that 72% of finance leaders had overspent on AI initiatives over the prior year. Today, as Esker looks ahead to 2027, the company’s employee-centric cost-allocation model serves as a pioneering case study in corporate financial control amid the generative AI boom.

Supporting Data and Industry Context

Esker’s internal struggles with budget overruns are indicative of a macroeconomic trend reshaping how finance departments view artificial intelligence. A constellation of recent studies and surveys highlights the friction between aggressive AI adoption and traditional financial oversight.

The Scale of Overspending

According to an Esker-released survey from September 2026, 72% of finance leaders admitted to spending more than initially planned on artificial intelligence initiatives over the preceding 12 months. More concerning for executive boards, 65% of CFOs reported a severe difficulty in directly connecting their surging AI usage with concrete, measurable business outcomes.

The Looming Threat of Agentic AI

While basic AI assistants and chat interfaces have already stretched corporate budgets, the technological horizon threatens to exacerbate the problem exponentially. A recent research report published by Futurum Research in partnership with AI infrastructure provider QumulusAI sheds light on the next phase of enterprise AI: agentic systems.

Unlike passive assistants that respond to discrete prompts, autonomous AI agents execute complex, multi-step workflows independently. According to the Futurum report, agentic AI can multiply token consumption per task by up to 100 times. As organizations scale these autonomous systems, traditional per-token pricing models threaten to transform from a minor line-item variance into a critical barrier to cost control.

Departmental Consumption Breakdown

At Esker, the data revealed stark variances in how different operational units consume AI resources:

• R&D and Finance: Identified as the primary drivers of token overages. These teams utilize advanced code-generation, data-modeling, and analytical tools that inherently demand heavy token throughput.
• Other Functions: While customer service and marketing utilize AI extensively, their consumption patterns have proven more stable and easier to forecast relative to engineering and financial modeling operations.

Official Responses and Strategic Insights

In interviews discussing the shift in financial strategy, CFO Scott McDermott has been candid about the addictive nature of the technology and the unique challenges it presents to corporate accounting.

"As the models have gotten smarter, and the technology has become more addictive, companies’ reliance on it has grown and that’s leading to more and more costs," McDermott noted.

Reimagining the P&L Through Employee Costs

To combat the unpredictability of traditional IT expense reporting—where AI costs would hit the profit and loss (P&L) statement in erratic, volatile bursts—Esker pivoted to a proactive tracking model.

The process begins by ingesting actual usage data from AI vendors. The finance department then calculates a rolling run-rate AI cost per employee at the close of each month. By factoring in assumptions regarding future productivity changes and departmental scaling, Esker treats AI tokens much like standard human capital expenses.

"We start with actual usage data from our AI providers to estimate how much employees in each function are consuming," McDermott explained. "The approach gives Esker a more consistent view of its total employee costs, rather than relying on when AI expenses hit the profit and loss statement."

Balancing Investment with Growth

Beyond merely tracking costs, McDermott emphasizes the ultimate goal of corporate finance: ensuring that every dollar spent yields an adequate return on investment (ROI). While measuring the direct productivity gains of an employee using AI can be straightforward—such as time saved on coding or document drafting—quantifying the subsequent top-line revenue growth remains a complex puzzle.

"We’re certainly trying to strike a balance between investment and growth," McDermott said. "It’s probably one of the biggest challenges that I’ve faced in my career, but I’m pretty confident that we’re on the right track for 2027."

Broader Implications for Corporate Finance

Esker’s integration of AI token consumption into employee expense planning signals a potential paradigm shift for corporate governance across the technology sector and beyond.

1. The Death of Static IT Budgets

For decades, software licensing followed predictable SaaS (Software-as-a-Service) models: a fixed seat fee per user per month. Generative AI has shattered this model. Because token consumption fluctuates based on project intensity, model complexity, and worker habits, static IT budgets are no longer viable. Finance teams will increasingly need to adopt dynamic, usage-sensitive forecasting methods akin to utility billing or operational manufacturing costs.

2. Redefining Employee Value and Productivity Metrics

By tying AI costs directly to individual roles, companies like Esker are laying the groundwork for a new era of performance analytics. If an R&D engineer or financial analyst incurs significantly higher AI token costs than their peers, management will naturally evaluate whether those costs correlate with proportionally higher output, faster project completion, or superior code quality. This could eventually lead to role-based AI budgets or customized token allocations per job description.

3. Pressure on AI Vendors to Innovate Pricing

The friction caused by per-token pricing models is unlikely to go unnoticed by AI infrastructure providers. As enterprise CFOs push back against unpredictable monthly bills and budget overruns, market demand will likely pressure vendors to develop more stable, predictable pricing structures—such as enterprise-wide flat rates for specific agentic workflows or guaranteed capacity tiers.

Until the market stabilizes, however, finance leaders will have to follow Esker’s lead: treating artificial intelligence not as a passive software tool, but as a dynamic, highly variable operational resource that must be closely monitored, forecasted, and integrated directly into the human capital equation.