If you want to calculate the ROI of an AI project, you need to consider more than just the licensing costs of the tool you’re using. The formula itself is simple. The challenge lies in determining exactly what factors should be included in the cost and benefit calculations. This is precisely where many business cases for AI projects paint an overly optimistic picture.
The ROI for an AI project is calculated as total benefits minus total costs, divided by total costs, multiplied by 100. On the cost side, most business cases fail to account for the internal expenses ranging from 20 to 35 percent and the follow-up costs after the launch. On the benefits side, these include time saved (calculated at the full cost rate), avoided error costs, and shorter turnaround times.
Why AI Projects Pay for Themselves Faster Than Traditional IT Projects
One reason for the interest in AI investments is the comparatively short payback period. For AI-driven automation, it typically ranges from three to six months, whereas for traditional IT projects, it ranges from 18 to 24 months (Sophera Consulting, 03/2026). In document processing, ROI figures ranging from 300 to 600 percent are reported after twelve months (consultingrechner.de, 08/2026).
For decision-makers in industry and retail, the question is not an academic exercise. An AI project competes for budget with a new plant or a warehouse extension, and those investments are traditionally costed soberly, with payback periods, risk discounts and a comparison of several options. An AI project ROI that stands up to the same sobriety convinces a management team more than an optimistic slide.
The comparison with traditional IT projects falls short in one respect. An AI project can often be tested in smaller steps before the full amount is invested. This ability to determine early on whether a use case works is a value in itself that is not reflected in any of the percentages cited.
The Cost Item That the AI Project ROI Usually Overlooks
of total costs go to internal effort: project management, specialist departments supplying test data, IT staff building interfaces. It is the most frequently overlooked item in AI business cases.
AI Consulting Germany, June 2026
| Item | Percentage | This is usually included in the business case |
|---|---|---|
| External Consulting and Development | 35 to 55 percent | Yes, there is an invoice. |
| Internal Expenses | 20 to 35 percent | no |
| Licenses and Operations | ongoing | in part |
| Maintenance of Data Sources After Launch | ongoing | no |
Anyone who fails to factor in the cost of their own employees’ time—because they’re there anyway—is artificially underestimating the value of their own investment. A realistic business case assigns an hourly rate to every hour of internal work, even if there is no invoice to back it up.
A second, often overlooked expense is the ongoing costs after the project is completed. Once a system is up and running, it still requires ongoing maintenance of data sources, occasional adjustments to accommodate changes in processes, and a point of contact for troubleshooting. Those who fail to factor these ongoing costs into their budget from the start will see a positive ROI in the first year but be surprised by unexpected bills in the second year.
What Really Matters on the Benefits Side
On the benefits side, several factors can be distinguished: saved labor time, avoided error costs, and faster turnaround times, which enable additional orders. Time savings can be converted into euros by multiplying the hours saved by the full cost rate for the respective employee group—not just by the gross salary. Social security contributions, workplace costs, and management expenses should be included in this calculation.
A pilot project should reach its initial break-even point after three to six months of live operation (AI Consulting Germany, 06/2026). If this point is not reached, that is no reason for alarm. It is a signal to review the use case. Some projects require more time because data preparation takes longer than planned. Others should be terminated at this stage because the original use case was too small to justify the investment. Monitoring this metric monthly makes this distinction apparent early on.
When an AI Project Should Be Discontinued
Every business case should include a clear criteria for termination before the project begins. If, after six months of live operation, no measurable benefits are evident and there is no plausible explanation for the delay, it’s better to stop the project than to wait another quarter.
In practice, however, it’s still difficult to call off a project because the money and time already invested make the decision emotionally charged. A project team that has spent three months working on an integration is reluctant to let that work go to waste. That is precisely why the criteria for termination should be documented in writing and reviewed by someone outside the project team; otherwise, the effort already expended automatically carries more weight than the cold, hard numbers.
The three pilot projects described in this article demonstrate how such a metric can be established as early as the project selection phase Generative AI in industry and retail.
Have the business case reviewedSend us your cost estimate. We'll let you know which cost items are missing and whether the expected benefits are realistically estimated.
Frequently Asked Questions
What types of benefits are factored into the ROI of an AI project?
Time savings, avoided error costs, and additional revenue from faster turnaround times are the most common types of benefits. Less tangible effects, such as higher employee satisfaction, are difficult to quantify and are better suited to a qualitative description than to the metric itself. Anyone who still wants to factor in such soft effects should report them separately so that it remains clear what portion of the result is based on hard numbers.
How do you quantify the value in euros of time saved at work?
The full cost rate for each employee group is the correct benchmark, not just the gross salary. Social security contributions, workplace costs, and a portion of management costs should be included in this rate; otherwise, the savings will be systematically underestimated.
When should an AI project be halted?
If, after three to six months of live operation, there is no sign of breaking even and no plausible explanation for the delay. A predefined termination criterion prevents a project from continuing longer than its numbers justify, simply for the sake of prestige.
The Next Step
torck costs AI projects for industry and retail using the same cost items described in this article, including internal effort and follow-on costs. Our teams in Maxhütte-Haidhof, Vienna and Rabat then implement the projects and keep running them. In an initial consultation we work through the AI project ROI for your own plan together. Schedule an Initial Consultation.
German 
English 
French