AI ROI
Why most AI investment still shows no return
ROI on AI is not hard to calculate because the maths is hard. It is hard because most companies never defined what they were measuring before they started spending.
Nineteen percent of Portuguese CEOs say AI increased their costs. Almost none can say by how much, or why.
A question almost nobody can answer
Ask a management team what AI returned this year and watch the silence. They can list subscriptions. They can name pilots. The euros are somewhere else — nobody tracked them.
That silence is the real cost. Not the subscription fee, but the twelve months spent unable to say whether the investment was right.
The highest cost of AI is not the subscription. It's the twelve months spent unable to say whether it was worth it.
The costs that never appear on the invoice
The nineteen percent reporting higher costs are not paying more for software. They are paying for integration nobody planned, correction nobody budgeted, and hours spent supervising a system that was never designed to be supervised.
A pilot that never leaves pilot stage is not free. It occupies attention, headcount and credibility that a defined project would have used better.
What to measure before spending, not after
A baseline: what the process costs today, in time or money, without AI. A single metric that would not move for any other reason. And a name — one person accountable for reading that number in six months.
Companies that do this before buying anything are the ones who, a year later, can actually answer the question this article opened with.