How to measure the ROI of intelligent automation and AI projects

Measuring the ROI of an automation or AI project is not just about comparing implementation costs with hours saved. It is about being able to answer, with numbers, the questions asked by those who approve the budget: what return can this investment generate, and how long will it take?
To structure this answer, it is necessary to identify the benefits, map the costs, and track what actually happens after the solution goes into production. One way to organize this analysis is to look at four dimensions: cost, reliability, capacity, and cycle.
Four dimensions for analyzing automation ROI
Cost. Reduction in expenses or costs avoided through automation. The manual work hours freed up, multiplied by the hourly cost of the team involved, help estimate their economic value. However, this result does not automatically represent financial savings: it is necessary to demonstrate how the freed-up capacity was used or which expense was avoided.
Reliability. Reduction in errors and rework. Each error corrected manually may involve time spent identifying and correcting it, as well as a potential impact on the customer. These costs should be measured to estimate how much automation can reduce them.
Capacity. The ability to absorb a higher volume of work with the available structure. This gain may avoid future expansion costs or allow the team to focus on other activities. To be included in financial ROI, its impact must be demonstrated using verifiable assumptions.
Cycle. The speed at which a process is completed end to end. Faster processes can contribute to business metrics such as conversion, customer response time, and deadline compliance. Converting this gain into financial value depends on demonstrating its relationship with business outcomes.
Not every project generates significant gains across all four dimensions at the same time. The first step is to identify where the process has the greatest potential for returns and document its current state: volume, execution time, errors, costs, and service level.
Double counting must also be avoided. If the same freed-up hours support both a capacity calculation and an avoided-cost calculation, their value should be included only once in the financial total.
What should be included in the cost calculation
Before calculating the return, it is necessary to map the full investment: assessment, development, licenses, infrastructure, integration with existing systems, training, and maintenance after the solution goes into production.
AI projects also include costs related to model usage, data preparation, quality evaluations, monitoring, and human review. Projects that account only for the initial implementation tend to overestimate ROI.
Benefits and costs should cover the same analysis period, such as the first 12 months of the project, including implementation, stabilization, and operation.
The basic formula
ROI (%) = [(Financial benefits − Total cost) ÷ Total cost] × 100
Financial benefits are the sum of the demonstrable monetary gains from the dimensions applicable to the process, without overlap. The analysis should distinguish between actual savings, avoided costs, and the estimated value of freed-up capacity.
The clearer the assumptions and evidence, the more robust the calculation presented to budget approvers will be. Benefits that have not yet been monetized can be tracked through operational metrics without forcing their inclusion in financial ROI.
The payback period should be calculated separately: it indicates when the accumulated net financial benefits offset the investment made.
Illustrative example, using fictitious figures: consider an automation project with the following hypothetical values for the first 12 months:
Item | Amount |
(+) Implementation and integration | R$ 60,000 |
(+) Licenses, infrastructure, and maintenance during the period | R$ 40,000 |
= Total cost | R$ 100,000 |
(-) Verified reduction in expenses related to outsourced processing | R$ 120,000 |
(-) Reduction in expenses related to corrections and reprocessing, not included in the previous item | R$ 30,000 |
= Total financial benefits | R$ 150,000 |
Applying the formula:
ROI = [(R$ 150,000 − R$ 100,000) ÷ R$ 100,000] × 100 = 50%
In this scenario, the payback period is approximately 6.5 months, considering the rate at which benefits accumulate throughout the first year.
How to sustain returns after implementation
As discussed in the Beyond AI ebook, individual gains in time, quality, and capacity are evidence of value. To sustain these returns, it is necessary to monitor the solution’s performance in production and the costs required to keep it running.
This includes metrics such as cost per transaction completed to the required quality standard, error and rework rates, the need for human intervention, and cycle time. In applications involving AI agents, it is also important to measure how many activities are completed without human intervention while remaining within the defined quality parameters.
A solution may free up many hours while also requiring more review or maintenance than expected. Continuous monitoring makes it possible to identify these deviations and update the estimated return based on actual results.
What this looks like in practice
At Ball, the hyperautomation journey has achieved ROI of up to 537% and savings of up to US$2.6 million. The case demonstrates the importance of tracking financial results alongside the governance, support, and operational evolution framework.
At FS Bioenergia, the Center of Excellence (CoE) journey combines automation, artificial intelligence, and workforce enablement to transform corporate processes. Tracking investments and the benefits generated indicates an ROI of 160%, considering gains calculated over a 12-month period as well as extraordinary gains. The case shows how measuring freed-up hours, avoided costs, and financial results helps guide the evolution of operations.
At Bridgestone, the partnership encompasses more than 200 automated processes across more than 25 areas and seven countries. Among the applications presented is the update of more than one thousand orders in SAP, completed in two days. The example demonstrates speed and responsiveness; to calculate its ROI, the result must be compared with the effort and cost of the manual alternative.
These indicators represent different dimensions of value. ROI percentage, financial savings, cost per transaction, and execution time complement one another, but they are not equivalent measures.
Smarthis’ experience
Over eight years, with more than 120 companies served across 25 countries, Smarthis has delivered more than 4,500 automations and returned more than 2 million hours to its clients’ businesses. This experience reinforces the importance of connecting technical metrics to the objectives of each operation, supported by governance and ongoing performance monitoring.
Next step
Analyzing cost, reliability, capacity, and cycle helps identify benefits that might otherwise be left out of the calculation. The calculation needs to consider which of these translate into financial results, which represent freed-up capacity, and which should be tracked through operational metrics.
Talk to a specialist at Smarthis to map the potential ROI of your process and define how to track results after implementation.





