AI ROI Is Mostly a Labor Question

There is a lot of discussion about the return on investment from artificial intelligence, but most of it starts from the wrong place. Companies look at what they are spending on models, software, infrastructure, consultants, and internal teams, then try to identify revenue generation or cost savings that can be directly attributed to those investments. Sometimes that works, but it often fails to capture the unique aspects of AI, because AI is not merely another category of software. Much of its economic value comes from changing what people can produce, how quickly they can produce it, and how organizations deploy talent against a given level of output. That makes AI ROI, at least in the near term, largely a labor question.

AI’s Real Impact

Imagine a company with ten engineers. Before AI, the company expected to add several more as the product and customer base grew, assuming no change in tools. But better development tools may allow the existing team to absorb more of that work while also changing the profile of the people the company hires next. The result could mean less hiring, but not necessarily. It may be different hiring, with some work that previously required additional execution capacity being absorbed by existing teams while the company places a higher premium on seniority, judgment, architecture, domain expertise, or people who can operate across a broader surface area. This shift in hiring – who is hired and their impact on the business – is an important form of AI return on investment.

Businesses usually make labor decisions at the margin. They ask how much they expect to grow, what capacity that growth will require, and what mix of talent, technology, and process is necessary to support it. AI changes those calculations, which is why the relevant productivity question is not whether an employee can perform a specific task 30 percent faster. A company is more than the sum of its developers’ tasks. If an engineer writes code more quickly but the organization cannot convert that time into better products, greater throughput, or higher-quality decisions, the business may capture very little value.

The key question is whether the organization can produce more, move faster, or take on more ambitious work without increasing every input proportionately. And the benefits of these tools extend beyond the core teams building software products. Yes, a company may ship a product earlier, but it may also mean a customer support team can handle more accounts, a marketing organization can produce more with the same staff, an engineering team can control a wider technical surface, or a manager can spend less time synthesizing information and more time making decisions. In each case, AI changes the economics of how work is organized.

Sources of ROI

In short, a more useful framework would consider at least four sources of return: direct revenue, direct cost savings, time compression, and increased organizational capacity. This is why I am skeptical of AI ROI calculations built primarily around tool costs. Whether a company spends $30 or $100 per employee per month on an AI product is usually not the important number. If the technology meaningfully changes organizational capacity, the economic effect should be much larger than the software bill.

The last two are particularly easy to underestimate because they often appear indirectly. A project that took three months instead of six does not necessarily create a new line on the income statement. A team that can take on an additional product line without immediately creating an entirely new function may not show an obvious accounting gain. Yet both can materially change the business’s economics.

Hiring Impacts

There is another consequence. If AI raises the productivity of existing workers, it changes the threshold for what constitutes a worthwhile hire. That does not mean companies stop hiring; it means the minimum viable hire changes. And the impacts and benefits will vary by situation: Strong people may become more valuable because they can control more surface area with the assistance of AI. Generalists may become capable of work that once required narrower specialization, while specialists may become more valuable where judgment, technical depth, accountability, or domain expertise still matter enormously.

The mistake is to reduce this to whether AI “replaces jobs.” The more immediate effect may be that AI changes team composition, the capabilities companies value, and the amount of output each combination of people and technology can produce. That is where I would start when calculating AI ROI. Do not ask only whether the technology produced identifiable revenue or eliminated an existing expense. Ask whether the organization became more capable, because if revenue, projects, customers, quality, and technical ambition can grow faster than organizational complexity, that gap may be the most important AI return of all.