How CFOs Should Think About the Balance Sheet
AI adoption is often framed as a technology decision or an accounting question. It is neither. It is a capital allocation decision.
The real issue for finance leaders is not whether AI should be expensed or capitalized, but whether it is being deployed as consumption or as capacity. Capital creates durable capability. Expense funds activity that resets each period. AI now does both.
Early deployments looked like tools – assistants, automation layers, productivity enhancements. They increased output without changing how work was organized and were appropriately treated as operating expenses. What has changed is where AI is now being embedded: inside core workflows, decision paths, and control functions historically defined by labor and judgment.
In those contexts, AI is no longer a tool employees use. It is the operating capacity the business relies on.
This is why disciplined CFOs are not focused on whether AI is “CapEx-eligible.” They are asking whether the capability is durable, whether it alters cost structure or decision velocity in a lasting way, and whether removing it would materially disrupt operations. If AI disappears when the subscription ends, it behaves like an expense. If the business cannot function the same without it, it behaves like capital.
In practice, AI spend is rarely singular. There is typically a build component – design, integration, workflow creation, and a run component – platform access, usage, monitoring. Treating both the same obscures the economics. Separating them allows finance teams to align spend with how value is actually created.
Across enterprise deployments, three structures tend to emerge. In some cases, AI remains an operating expense, optimized for speed and flexibility. In others, it becomes operating infrastructure, embedded into financial, operational, or compliance systems, where building capability is distinct from running it. In a smaller number of cases, AI is tied directly to revenue or margin expansion and structured through joint development or revenue-sharing arrangements.
At Performance AI, we’ve seen all three models coexist within the same organization, applied deliberately based on risk tolerance, time horizon, and strategic importance. The distinction is not the sophistication of the AI, but whether it changes how the business operates at scale.
This moment mirrors earlier shifts: from on-prem to cloud, from labor-driven models to systems-driven ones. Each required CFOs to rethink how investment showed up on the balance sheet. AI is the next iteration, arriving faster and touching more of the organization at once.
The task for finance leaders is not to force AI into OpEx or CapEx prematurely, but to structure AI spend so the balance sheet reflects durable operating capability rather than recurring activity. When that is done well, accounting treatment follows.
AI becomes capital not when it is capitalized, but when it changes how the company works.
