There’s a story that a well-known HR software technology company spent a very high amount on artificial intelligence in just a few months and that, faced with a lack of clear results, it ended up building an internal tool to measure what each employee actually contributed with those new capabilities
At Tisa, we have been helping companies leverage technology with a business rationale—not because it’s trendy—since 1987. And when it comes to AI, our starting point is always the same: first the problem and the metric, then the tool.
Why so many companies spend on AI without a return
The pattern repeats itself. A company hears about artificial intelligence, buys licenses, tries out several tools and, after a few months, someone asks at a management committee meeting: «And what has this brought us?». The silence that follows is usually due to three prior mistakes.
- Buying technology before defining the problem. «We want AI» is not an objective; «cutting the response time on a quote in half» is.
- Not setting the success metric at the beginning. If you don’t know what you were going to improve, you won’t be able to prove you’ve improved it.
- Ignoring the data. Without organized and governed data, AI doesn’t perform: it repeats errors, «hallucinates» and generates distrust.
The reaction of creating a tool to measure impact comes late, but it points in the right direction. The lesson for a small or medium-sized company is clear: that measurement is not an extra added at the end, but the scaffolding that supports the entire project from minute one.
What it really means to measure the ROI of AI
Measuring the return of an artificial intelligence project is not about counting how many licenses you’ve activated or how many employees «use AI». Those are vanity metrics. The ROI of AI is calculated by comparing what you invest with the business value you obtain, and that value must be observable.
Costs you must add up
- Licenses and platform consumption (remember that many models bill by usage).
- Hours of development, integration and fine-tuning of the use case.
- Data preparation and governance.
- Team training and adoption time.
- Subsequent maintenance and supervision.
Benefits you must quantify
- Time saved on repetitive tasks (handling inquiries, drafting documents, classifying emails).
- Errors avoided and their associated cost (rework, complaints, penalties).
- Revenue influenced: faster responses to customers, quotes delivered sooner, fewer lost opportunities.
- Capacity freed up: people who stop doing mechanical work and devote themselves to higher-value tasks.
The key is to record the baseline before starting. If today responding to an order takes an average of 40 minutes, that number is your starting point. Without it, any future improvement will be an opinion, not a data point.
Start with a measurable use case, not with the tool
Our practical recommendation is simple: instead of buying AI «for everything», choose one use case tailored to the business with a measurable result and narrow it down. A good use case meets four conditions: it hurts enough to justify the effort, it happens frequently, it has a clear metric and it has data available.
Some examples any private company will recognize:
- An assistant that answers internal inquiries about procedures or documentation, supported by the company’s own knowledge base (using the RAG technique) to avoid made-up answers.
- The automatic classification and prioritization of incoming emails or incidents.
- The extraction of data from invoices or delivery notes to load them into the ERP without typing them by hand.
- The generation of drafts of offers or replies that a person reviews and approves.
Each of these cases has an obvious metric: minutes per operation, accuracy percentage, number of documents processed per day. That is exactly what makes it possible to honestly measure the ROI of AI.
The role of data and Power Platform
This is where custom development makes the difference. AI does not live in isolation: it needs to connect to your systems and your data. With Microsoft Power Platform —Power Automate to automate processes, Power Apps to create custom applications and Power BI to visualize results— you can build both the use case and the dashboard that measures its impact.
Imagine a flow in Power Automate that processes incoming invoices with AI and records, on each run, how much time it has saved and with what degree of accuracy. That record feeds a Power BI dashboard that management consults whenever it wants. That way, the question «what has AI brought us?» stops being scary: the answer is on screen, updated and with real data.
All of this rests on an essential foundation: data governance. Defining what data is used, who accesses it and how its quality is guaranteed is what prevents a promising project from ending up generating distrust. Without reliable data, there is no reliable AI or ROI worth talking about.
How we approach it at Tisa
Our approach to AI applied to business is phased and adapts to the maturity of each company. We start with an assessment that identifies the relevant data, evaluates the tools according to your investment capacity and defines, together with you, the first use case with its metric. From there we develop and execute that case, and we train your team to make the most of it. The goal is not «to have AI»: it’s for the investment to translate into fewer costs, fewer errors and more visibility.
As a Microsoft partner and ISV with more than 30 years of experience, we combine business knowledge with the right technology, without selling you smoke or promising what cannot be proven.
Conclusion: measure first, invest later
The moral of the case that opened this article is not «AI is useless». It’s that measuring the ROI of AI cannot be an afterthought, but rather the starting point. Define the problem, set the metric, prepare the data, choose a narrowed-down use case and build the dashboard that will tell you, with data, whether it’s worth it. That way AI stops being an uncertain expense and becomes an investment you can defend before any committee.
Do you want to identify your first AI use case with a measurable return? At Tisa we help you with a no-obligation assessment. Write to us at info@grupotisa.com, call us at (+34) 971 305 885 or visit grupotisa.com.