The following is a guest article by Daniel Schmeltz, Managing Director of Enterprise Performance Improvement at Alvarez & Marsal. The views expressed are solely those of the author.

Companies are investing heavily in artificial intelligence, but few can demonstrate sustained financial returns. Despite widespread experimentation and deployment, AI rarely translates into measurable improvements in EBITDA, cash flow, or return on invested capital—and as global executives shift theirfocus from AI adoption to AI return on investment, this concern is becoming increasingly prominent.

The problem is not the technology itself, but that the introduction of AI has not been accompanied by the operating model transformation needed to convert capabilities into value. Unless AI is embedded in the processes of decision-making, accountability, and governance, it will continue to generate visible activity while struggling to leave a lasting impact.

The gap between investment and impact is clearly visible in corporate data. Arecent MIT study on AI adoptionfound that despite a surge in global investment, most organizations have not yet seen measurable financial results. While this may oversimplify the value organizations have realized from AI investments to date, it reveals a macro reality: the true financial benefits organizations are currently deriving from their AI investments are minimal.

This disconnect is evident in the way projects are advanced: multiple pilots are launched simultaneously across functions and business units, demonstrating activity but rarely achievingscale. Few pilots are tied to financial owners, baselines, or measurable economic increments, making it impossible to determine success or failure.

This pattern stems fromurgency rather than design; leaders feel pressure to show progress and avoid falling behind competitors, so numerous departments are asked to deploy AI without a clear definition of success.

As a result, AI dominates the operating model rather than supporting it. AI is deployed in "HR" or "Finance" departments, and this vague mandate leaves companies wondering after pilots end: "What next?" AI can make hiring decisions or prepare budgets, but who owns the decision—the manager or the machine? And when decisions involve more sensitive information, how are control mechanisms maintained?

An operating model is defined by who makes decisions, how work flows, and where accountability lies. Most AI deployments overlook this, treating the technology merely as a simple automation tool.

Instead, companies should view AI as a team of enthusiastic but inexperienced "digital interns," not as software. They have boundless energy and can process data at scale, but they lack judgment. A successful operating model does not replace humans with these "digital interns" but elevates humans to editors, responsible for curating and validating the interns' work. In this way, technical capability is combined with operational authority to ensure lasting impact.

The key is that this shift not only improves efficiency but also expands customer value. When "digital interns" take on repetitive, low-value work, humans are freed to focus on higher-order creative and judgment tasks. In this model, AI becomes a true thought partner—helping teams design better products and personalize experiences—enabling organizations to create entirely new revenue streams rather than merely running the same operating model at lower cost.

Evidence from early enterprise deployments points to a more effective approach. The strongestresultscome fromhuman-led, AI-augmented models. Rather than replacing human decision-makers with AI models (as financial institutions do when automating credit and risk decisions), AI should be used in the decision preparation phase.

This distinction is critical in high-stakes decisions—what might be called the "cabinet test." If an AI-generated cabinet image is slightly off, the cost is zero; if it miscalculates a credit risk covenant, the cost is existential. AI excels at pattern recognition, synthesis, and drafting, but humans must retain judgment responsibility in high-risk areas. Over time, as confidence grows and performance stabilizes, human intervention can be reduced to handling exceptions only, rather than continuous involvement. In practice, enterprise AI projects that deliver measurable financial returns often follow a consistent pattern:

  1. Anchor AI projects to financial value (cost, revenue, cash flow—not functions). This means applying the "cash flow litmus test": if a pilot cannot be tied to a specific financial metric (such as days sales outstanding or inventory turnover), it may be better suited for an R&D lab than an operating budget.
  2. Redesign end-to-end human-machine workflows before automating decisions.
  3. Invest in change management and training, building familiarity through actual use rather than relying on top-down mandates.
  4. Expand governance and automation scope only after trust and control are established.

This sequence is critical. Automating decisions too early introduces risk and erodes trust; starting with augmentation builds credibility, enabling subsequent expansion, as seen in areas such as software development, pricing analysis, and high-volume transaction processing.

Anotherconsistent findingis that impact improves when AI projects are aligned with value rather than organizational silos. More successful efforts begin with outcome orientation, such as AI applications focused on service cost or cycle time. When use cases are tied to economic levers, prioritization becomes clearer and trade-offs more explicit.

Change management also affects outcomes. AI adoption is won through repeated, low-risk use in real work, not through policies, training materials, or executive mandates. When early adopters demonstrate tangible value in daily work, skepticism fades and familiarity grows. Consistency, not intensity, is whatbuilds lasting operational capability。

Taken together, these patterns point to a clear conclusion: it is still premature to let AI run the enterprise operating model end-to-end. Instead, AI should become a key part of reshaping how work is done. Organizations making progress are deliberately integrating AI into people, processes, and technology, while keeping humans firmly in charge of critical decision accountability.

The companies that ultimately win the AI race will not be those that deploy the most tools, but those that treat AI like any other transformation investment—anchored in value, governed with discipline, and embedded in actual decision processes. AI can accelerate an operating model, but it cannot compensate for the absence of one.