AI projects: how context, data, and adoption contribute to returns
Luís Paravato
The return on an AI project depends on its use in day-to-day work, the quality of its outputs, and the benefits the company realizes. Purchasing a tool or completing a pilot is one step in that process.
To assess results, your company needs to understand the activity before implementation, the investment required, and the changes observed afterward. Data, people, processes, and maintenance all factor into that assessment.
A fast application that requires extensive review delivers a different result from what its generation speed suggests. Evaluation needs to cover the entire workflow, from information input to a deliverable accepted by the user.
Context determines what to measure
Every project starts with an activity and an expected outcome. An internal information assistant aims to provide faster access to accurate information. A customer service application aims to resolve requests effectively and route them appropriately.
These purposes require different measures. The number of responses generated indicates usage, but does not demonstrate usefulness or return.
The plan documents the baseline, users, volume, expected quality, and responsibilities. It also states which activities are outside the scope, allowing results to be interpreted without expanding the promise after implementation.
Establish a baseline for comparison
Before testing, track time, errors, rework, and task completion times. Record differences by case type and user experience.
Use comparable situations during evaluation. A trained team handling simple cases should not be compared directly with a team handling complex exceptions without documenting that difference.
Changes in demand, staffing, processes, and offerings belong in the analysis record. They help distinguish changes associated with the AI application from other changes during the period.
The project's total investment
| Component | What to consider |
|---|---|
| Implementation | Design, configuration, integration, and testing. |
| Data | Organization, updating, and processing of sources. |
| Technology | Licenses, usage, infrastructure, and connections. |
| People | Training, review, support, and management. |
| Maintenance | Fixes, new evaluations, and version changes. |
| Exceptions | Rework and routing of cases the application cannot handle. |
The analysis states the period covered. Initial investment and recurring expenses behave differently and need to be shown separately.
Avoiding double counting is also essential. The same time freed up should not be presented as both an expense reduction and additional capacity sold without evidence of both effects.
Freed-up capacity and financial return
Hours saved indicate an operational change. To estimate economic value, the company examines how that capacity will be used and what benefit that use produces.
A return on investment calculation uses verified incremental financial benefits and the total investment for the period. The formula and assumptions should accompany the presentation.
Quality benefits also deserve to be recorded, even when they cannot reliably be converted into monetary value. Fewer errors, better traceability, and more consistent service should be reported as outcomes in their own right, without invented financial values.
Adoption must include usage and quality
Tool access and the number of registered users indicate availability. Effective adoption occurs when the application is used for the intended tasks and delivers usable results.
Monitoring covers frequency, completed tasks, corrections, and workflow abandonment. Conversations with users explain access difficulties, incomplete sources, or steps that have added work.
Leaders need to provide training, time to learn, and clear ownership of adjustments. Requiring usage without fixing functional problems produces access logs without operational benefits.
Illustrative example: report preparation
A company tests AI to prepare first drafts of client reports. The pilot reduces writing time, but review reveals incorrect numbers and references to different periods.
The team organizes the sources, standardizes the analysis period, and limits generation to validated data. A new evaluation measures preparation, verification, and correction of the complete report.
The result now accounts for total time and accepted quality. The project also tracks source maintenance and training for the professionals who review the deliverables.
The freed-up capacity is directed toward additional analyses requested by clients. Its commercial impact is tracked separately, without automatically attributing every renewal to AI use.
When to expand, adjust, or discontinue
Conditions for continuing need to be defined before expansion. Minimum quality, operating investment, effective usage, and maintenance capacity guide the evaluation.
An application that benefits only a few cases retains a limited scope. Recurring failures require correction and another review. A lack of demonstrable value justifies discontinuing the application and documenting lessons learned.
After implementation, new sources and versions are evaluated before replacing components in use. Monitoring maintains enough history to identify changes in performance.
Frequently asked questions
How do you calculate the ROI of an AI project?
Define the period, incremental financial benefits, and total investment. Calculate the difference between benefits and investment, divided by investment, and present the assumptions alongside the result.
Does time saved equal an expense reduction?
No. A reduction depends on an actual financial change. Time freed up represents operational capacity until its use is demonstrated.
Does a successful pilot prove scalability?
A pilot evaluates the conditions tested. Expansion requires checking volume, case diversity, maintenance, and support for additional users.
How does Kronos Experience work on these projects?
Kronos connects the application, data, daily workflows, and business results, tracking implementation, usage, and economic contribution.
About Kronos Experience
Kronos Experience is a Brazilian business strategy and intelligence consultancy specializing in market, customer, product, and revenue intelligence for digital and service businesses.
We work to increase your company's value to the market and its customers, turning that value into a competitive advantage. Our work connects market strategy, positioning, brand, acquisition, paid media, and sales with customer intelligence, experience, product, data, retention, and monetization.
Through assessment, strategic direction, implementation, and ongoing monitoring, we structure opportunities to increase returns on your brand, customer base, channels, products, and infrastructure, while developing new offerings and revenue streams.
Written by Luís Paravato
