AI without ROI: how to avoid wasting 95% of projects

In brief: How to calculate ROI on an AI project

Most AI projects fail because economic due diligence is missing. Here's how to calculate ROI before investing:

Practical rule: If the payback period exceeds 18-24 months, the project probably isn't worth it.

Free tool: AI ROI Calculator – 5 minutes to know if your project makes economic sense.

In recent years AI has become the new "mandatory budget line". Many companies feel almost obligated to launch an artificial intelligence project, often without rigorous economic evaluation. The result is what various international studies are now showing with clarity: most initiatives don't generate measurable returns, or they do so in timeframes incompatible with board expectations.

The problem isn't the technology, but the absence of economic due diligence before starting. The decision is made "by gut feel", based on wow effect or competitive pressure ("competitors are doing it too"), rarely on numerical scenarios: investments, expected savings, productivity increases, payback times.

The AI paradox: investments without numbers

In any other investment - a new plant, a production line, management software - it's normal to talk about ROI, payback period, expected cash flows. With AI, though, we often settle for vague formulas: "we'll improve efficiency", "we'll reduce errors", "we'll automate low-value activities".

This approach creates two problems:

1. Projects that start without solid ground and are abandoned after months of work
2. Misaligned expectations between whoever proposes the project and whoever must approve the budget

The necessary step is to put numbers back at the center:

• Estimate time savings and translate them into labor cost
• Quantify error reduction (and related costs)
• Measure impact on productivity and ability to handle more volume with the same staff
• Connect these effects to an initial investment and realistic adoption curve

Only then does it make sense to talk about AI ROI, and compare different projects.

A real case: the invoice bot that didn't make sense

Recently a company contacted me with an apparently reasonable request: integrate an automatic bot for sending invoices to clients. The idea was simple - automate a repetitive process, free up time for the admin team, reduce sending errors.

Before starting with technical analysis and quotes, I launched what I call a process of economic interview: a series of structured questions to understand if the project, in the numbers, really made sense.

What emerged from the interview

During the analysis, the real dimensions of the problem became clear:

• Invoice volume: a few hundred per month
• Current time for manual sending: a few minutes per invoice, already partially automated
• Errors: rare, easily managed
• Maximum optimization estimated: about 40 hours/year

Translated into economic terms, considering average admin team RAL, annual savings came to around 1,000€.

The calculation that changed the decision

Against savings of 1,000€/year, investment to develop, test, integrate and maintain the bot would have required:

• Initial development: several thousand euros
• Gestional system integration: medium complexity
• Annual maintenance: recurring costs
• Team time for testing and adoption: non-negligible

The payback period? Over 5 years, in an optimistic scenario. In a realistic scenario, the project would never reach breakeven.

The final decision

The company chose not to proceed with the automatic bot. Not because AI doesn't work, but because that specific use case, in that context, didn't make economic sense.

This is exactly the function of due diligence: avoid falling in love with technologically elegant but economically unsustainable solutions.

The AI ROI Calculator

For this reason, at Castaldo Solutions we chose to take a very concrete step: put online, for free and without registration, an AI ROI Calculator.

The tool guides the user through an interactive survey and returns:

• Estimated ROI as a percentage
• Payback period (how long to reach breakeven)
• Net Present Value of the project (NPV)
• Clear breakdown of benefits: time savings, error reduction, productivity increase, operational efficiency improvement

Methodology based on real benchmarks

The logic isn't built "by feel", but relies on public benchmarks from large consulting and technology players, adapted to typical contexts of sales, CRM and process automation.

At the end the decision maker can download a PDF report and use it as a basis for internal discussion with CFO, IT and business functions.

It's not a dream configurator, it's a filter: in some cases the result will suggest proceeding, in others it will show that the project, as conceived, doesn't hold up.

When AI makes sense (and when it doesn't)

An ex ante evaluation tool serves to:

✅ Proceed with projects that make sense

• Expected savings significant compared to investment
• Payback period compatible with company horizons
• Quantifiable and measurable benefits over time

❌ Avoid projects destined to fail

• Use cases with volumes too low to justify investment
• Already sufficiently optimized processes
• Benefits mainly "cosmetic" or image-related
• Integration complexity disproportionate to generated value

The right question before every AI project

The objective, ultimately, is simple: before saying 'let's do an AI project', ask yourself 'does this project, in the numbers, really make sense?'

If the answer is yes, AI stops being an image experiment and becomes finally an investment like any other: with risks, but also with an ROI you can explain, defend, and measure over time.

If the answer is no, better to know before spending months and budget on a project destined to be abandoned or, worse, to generate frustration and distrust toward AI in general.

👉 Evaluate your next AI project: Free ROI Calculator

Five minutes of analysis today could save you months of investments with no return. Start from the numbers, not from enthusiasm.