{"id":33656,"date":"2026-10-06T17:14:08","date_gmt":"2026-10-06T17:14:08","guid":{"rendered":"https:\/\/thecloud.group\/?p=33656"},"modified":"2026-10-06T17:23:02","modified_gmt":"2026-10-06T17:23:02","slug":"how-to-measure-roi-in-artificial-intelligence","status":"publish","type":"post","link":"https:\/\/thecloud.group\/en\/como-medir-roi-inteligencia-artificial\/","title":{"rendered":"The ROI of AI is not measured in prompts"},"content":{"rendered":"
\n\t\t\t\t\t\t
\n\t\t\t\t\t\t\t
<\/div>\n\t\t\t\t\t\t
\n\t\t\t\n\t\n<\/svg>\t\t<\/div>\n\t\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\"logo\"\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t

The ROI of AI is not measured in prompts\n\n<\/h1>\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t

October 6, 2026<\/h1>\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t
\n\t\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t

The number of queries, active users, or executed prompts doesn't measure the return on investment in artificial intelligence: it measures its use. These are different things, and confusing them is why so many projects are presented as successful to the committee while the bottom line shows nothing.<\/span><\/p>

A system with high adoption and zero return is perfectly possible. In fact, it's the most common scenario.<\/span><\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t

\n\t\t\t\t
\n\t\t\t\t\t

Why usage metrics are misleading\n<\/h2>\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t

A usage metric answers the question, "Are people using it?" A return on investment metric answers, "Is the company making a profit?" There's a gap between the two that's almost never documented.<\/span><\/p>

The typical case: a tool saves eighty people fifteen minutes a day. On paper, that's twenty workdays a month. In practice, it's fifteen minutes spread out and absorbed into other tasks, without any additional budget line item appearing.<\/span><\/p>

McKinsey quantified the phenomenon in its 2025 survey of 1,993 participants in 105 countries: 88.1% of organizations use AI in some function, but only 39.1% report an impact on EBIT, and barely 6.1% exceed 5.1% of attributable EBIT. High usage, rare return.<\/span><\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t

\n\t\t\t\t
\n\t\t\t\t\t

The three ways in which saving does lead to results\n<\/h2>\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t

Any serious business case has to explain which of these three it is pursuing:<\/span><\/p>

A whole step disappears.<\/b> A task isn't accelerated: it ceases to exist. A validation that is no longer needed because the data is checked at the source.<\/span><\/p>

More volume is absorbed without expanding the structure.<\/b> The team manages one more 30% case with the same staff. This is the most common scenario in growing companies and the easiest to defend.<\/span><\/p>

An identifiable direct cost is reduced.<\/b> Fewer errors, less rework, fewer contractual penalties, less overtime at month-end closing.<\/span><\/p>

If a project cannot specify which of the three it is seeking, it does not have a business case. It has a reasonable expectation, which is something else entirely.<\/span><\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t

\n\t\t\t\t
\n\t\t\t\t\t

The four metrics that a committee should require\n\n<\/h2>\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t

Metrics<\/b><\/p><\/th>

What does it measure?<\/b><\/p><\/th>

Why it matters<\/b><\/p><\/th><\/tr><\/thead>

Cycle time<\/b><\/p><\/td>

From the moment a case is entered until it is closed<\/span><\/p><\/td>

It reflects the entire process, not just the task.<\/span><\/p><\/td><\/tr>

Cost per case resolved<\/b><\/p><\/td>

Total cost divided by closed cases<\/span><\/p><\/td>

It is comparable to the previous process<\/span><\/p><\/td><\/tr>

Error or rework rate<\/b><\/p><\/td>

Cases that need to be corrected afterwards<\/span><\/p><\/td>

Detects savings that are paid elsewhere<\/span><\/p><\/td><\/tr>

Absorbed capacity<\/b><\/p><\/td>

Volume managed with the same structure<\/span><\/p><\/td>

Translate savings into growth<\/span><\/p><\/td><\/tr><\/tbody><\/table>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t

\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t

The third option avoids the most expensive trap. A system that resolves cases for \u20ac0.40 compared to \u20ac3 for the manual process seems like a resounding success, until it's discovered that the 20% requires a subsequent correction of \u20ac5. The net result is different, and only becomes apparent if someone measures it.<\/span><\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t

\n\t\t\t\t
\n\t\t\t\t\t

The baseline: what to do before you start\n\n\n<\/h2>\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t

None of these metrics are useful without the previous value. And the time to capture that value is before making any changes, because afterward, it can't be reconstructed without discussion.<\/span><\/p>

Four data points, measured over two or three weeks:<\/span><\/p>

  1. <\/b>How many cases are processed<\/b> during the period.<\/span><\/li>
  2. <\/b>How long does each one take?<\/b>, including waiting and coordination, not just active work.<\/span><\/li>
  3. <\/b>How many require correction<\/b> later and what that correction costs.<\/span><\/li>
  4. <\/b>How much does the entire process cost?<\/b> For example, with the imputed personnel cost.<\/span><\/li><\/ol>

    This job has a valuable property: <\/span>It produces results even if the project is not carried out.<\/b>. Measuring a process reveals bottlenecks that are sometimes corrected without any technology.<\/span><\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t

    \n\t\t\t\t
    \n\t\t\t\t\t

    The decision threshold, written before\n\n<\/h2>\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
    \n\t\t\t\t
    \n\t\t\t\t\t\t\t\t\t

    The second discipline that separates successful projects from those that drag on forever is to set in advance the number that determines whether to continue or stop.<\/span><\/p>

    It must meet three conditions: be a business metric, have a specific value, and have a deadline. "If the cost per case doesn't fall below X in eight weeks, we'll stop" is a threshold. "We will continuously evaluate the impact" is not.<\/span><\/p>

    Without a written threshold, the evaluation only happens when too much has already been invested to admit that it's not working. This is the mechanism that produces the permanent pilot programs we described in <\/span>the pilot's purgatory<\/span><\/a><\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t

    \n\t\t\t\t
    \n\t\t\t\t\t

    The cost that almost no one includes\n\n<\/h2>\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
    \n\t\t\t\t
    \n\t\t\t\t\t\t\t\t\t

    An honest return calculation should subtract five items, not one:<\/span><\/p>