The Real Impact of AI on Business and the Cost Trap
Luiz D'Elboux analyzes the gap between AI spending, actual revenue, and token cost management.
Luiz D'Elboux discusses corporate AI adoption, warning of the mismatch between technical expenses and actual revenue growth. The debate draws lessons from case studies like Klarna and Uber, highlights Apple's cautious strategy, and explains how AI agent orchestration prevents budget waste.
Luiz D'Elboux, SaaS board advisor and former director at GoDaddy, spoke with Bruno Gimenes on Marketing - The Podcast about the actual maturity of artificial intelligence in business. In an environment where leaders view AI as an existential necessity, hasty decisions around layoffs and uncontrolled budget allocation have exposed significant operational and financial bottlenecks.
In the video above, D'Elboux breaks down how the unchecked race for automation triggered a cycle of misguided hiring cuts and unexpected expenses, highlighting what separates genuine efficiency gains from investments that yield no direct return.
Rash Decisions and the Rehiring Cycle
According to D'Elboux, executive pressure led many leaders to treat AI adoption with existential urgency: "If we don't start getting the hang of this, we'll be out of business in two years." This initial frenzy prompted companies to approve massive budgets for experimental tools without a clear strategic focus, leading to sweeping layoffs that quickly backfired.
This shift caused major operational disruptions. As D'Elboux explains: "Now, after a year and a half or two years of this madness, what do we see? A ton of companies having to rehire people. Why? Because AI couldn't fill the shoes of those employees, couldn't handle their tasks, or ended up doing so at a much higher cost." He points to Klarna, the payments company that laid off a large portion of its customer support staff in early 2024 to deploy chatbots. By focusing strictly on reducing wait times and queue lengths rather than customer retention, service quality plummeted, forcing the company to bring human teams back. In another instance of unbalanced spending, Uber reportedly burned through its entire projected annual AI budget in the first three months of 2026.
Marketing Productivity vs. Flat Revenue
While there are clear productivity gains in specific operational areas—such as lean teams of five or six people managing global campaigns and rapid content translation across monthly budgets in the hundreds of thousands of dollars—these internal efficiencies do not automatically translate into top-line revenue growth for traditional businesses.
D'Elboux emphasizes the divide between technology providers and end consumers: "The companies selling AI solutions and AI infrastructure are seeing massive revenue growth." Infrastructure providers, data centers, cloud platforms like Google Cloud, AWS, and Azure, as well as foundation model creators like OpenAI, are posting record numbers. In contrast, quarterly reports from publicly traded companies in legacy sectors show revenue trajectories remaining flat, reflecting higher capital expenditures without a corresponding increase in net sales.
Contrasting this aggressive spending spree, D'Elboux highlights Apple's measured stance. The tech giant has maintained conservative investments compared to Microsoft, Google, and Meta, sustaining its market growth without committing to massive data center capital expenditures.
The Token Trap and Agent Orchestration
The software pricing shift from flat-rate subscriptions to token consumption—a consumption-based model already seen across Salesforce platforms—demands strict cost governance to protect operating margins. D'Elboux references Scott Galloway's remarks regarding L'Oréal's massive Nvidia chip purchases, questioning whether consumer goods brands should be investing heavily in foundational AI infrastructure.
To curb waste, D'Elboux advocates for technical agent orchestration, matching model costs to the complexity of each specific workflow. Routine tasks do not require state-of-the-art models: "You wouldn't assign an agent to simply proofread ad copy with an advanced model; you'd use a lightweight model just to catch semantic errors." In his view, financial efficiency comes from reserving costly frontier models exclusively for high-leverage creative or strategic tasks, keeping overall operations tightly governed.