AI Delivers No Financial Payoff to Most Companies. How to Check It in Practice

What the economist actually claims

The starting point is simple and uncomfortable. According to TechCrunch, which cites an economist at Apollo, despite massive investment AI has delivered no financial payoff to corporate America outside the tech sector. The claim concerns profit, not enthusiasm or the number of pilots launched. For a management board, this means the question of whether a deployment makes sense must be answered with numbers, not at a conference.

How to measure the effect of a deployment

Before signing a contract, record the starting point: unit cost of the process, cycle time, error rate, or revenue from a given channel. Without this baseline, every result becomes a story. Then launch the deployment with a control group, meaning a part of the team or process that keeps working the old way. Only the difference between the groups shows whether AI changed anything.

The sequence of steps looks like this:

1. Choose one process with a measurable cost or revenue.
2. Measure the starting state over at least several weeks.
3. Compare the result with the control group after a fixed period.
4. Convert the result into money only after verification.

Where the pitfalls are

The most common mistake is counting saved hours as profit. If the time was not redirected to sales, and headcount or external costs did not fall, the saving is only theoretical. The second mistake is treating a pilot as a full deployment. Integration, maintenance and training costs appear only in production. The third pitfall is trusting metrics supplied by the vendor, which rarely reflect your data and your processes.

Decision criteria should be written down before launch:

1. a threshold below which the project is stopped,
2. an evaluation date, for example after two quarters,
3. an owner of the financial result on the business side, not in IT.

Obligations that do not wait for results

Regardless of payoff, in the EU Article 4 of the AI Act (Regulation (EU) 2024/1689) applies. Providers and deployers of AI systems must take measures to ensure a sufficient level of AI literacy among staff who operate and use these systems. This obligation has applied since 2 February 2025. Skipping training in pursuit of quick returns can therefore also be a formal error. It is worth building it into the deployment plan from day one.

Summary

The claim that AI does not pay off for most companies does not mean the technology is useless. It means that returns depend on measurement discipline, not on the mere presence of a tool. A company that starts with one process, a baseline and a control group will know whether its investment works before the result shows up in its financial statements.

Lub System helps B2B companies implement AI, automation and IT solutions end-to-end - from strategy to deployment. See our services or get in touch to discuss your case.