Artificial intelligence and real estate agencies: an Italian case, read with caution

Professionecasa, one of Italy's leading real estate franchises, announced a partnership with Metacasa, a platform bringing artificial intelligence into agency work, and published a set of numbers to describe the results: +6% on the average sale price, +161% in contacts from real estate portals, +237% in properties saved as favorites, +55% in sales of hard-to-place properties. These are striking numbers, and that's exactly why they're worth reading carefully before treating them as a benchmark.

What Professionecasa's announcement says

On the page announcing the partnership, Professionecasa presents these four numbers as "results achieved by agencies already using the platform," but also introduces them as "an indicator of the potential Metacasa offers": two different framings on the same page, one implying a measured outcome and one implying a projection. There's no note on how many agencies were involved, over what period the increases are calculated, or who carried out the measurement. No date appears on the page.

Why those numbers don't add up

Metacasa's own website has a section dedicated to results, introduced with the line "not estimates, not projections: field-verified performance." But the numbers it reports for comparable benefits are different from Professionecasa's: it cites +94% in clicks on listings, not +161% in contacts, and deals closing 40% faster. For the most directly comparable category, properties needing renovation or hard to sell, Metacasa's own site states a 30% increase, while Professionecasa, describing the same product, states 55%. Two sources tied to the same service report two different figures for the same result category, and neither states a shared methodology. No trade publication covering Italian proptech independently reports these four numbers: they appear only in the two companies' own promotional material.

This isn't a question of good faith: a company announcing a result chooses the most favorable figure among those available, not necessarily the most representative one, and it's rare for anyone to verify it from the outside. That's simply how a partnership announcement works by construction, a different logic from an independent measurement with a stated sample.

Even a widely cited digitalization figure has a calculation error

In 2022, several Italian outlets, including Il Sole 24 Ore, ran headlines claiming only 16% of Italian real estate agencies had started a digital transformation. The body of that same article, though, explains that the survey cited, based on a sample of 200 agents presented at a conference in Rome, counted 16 who had genuinely started such a path. Sixteen out of two hundred is 8%, not 16%: the headline turned an absolute count into a percentage identical to the numerator, and multiple outlets repeated the same figure without redoing the math. It's also a four-year-old figure, from a small sample with no published methodology: I'm not using it as a reference for today, but it's worth knowing, because it shows how a misreading can circulate for years even when the source isn't a company selling something.

The Italian real estate market really is growing

One figure that does hold up: in 2025 there were about 767,000 residential real estate transactions in Italy, up 6.5% from 2024, according to the FIAIP Monitora Italia 2025 report, produced by the FIAIP Research Center with ENEA and I-Com and presented at the Italian Senate on March 17, 2026. The number doesn't rest on a single interested source: the Italian Revenue Agency's Rapporto Immobiliare 2026, based on cadastral data and published independently a few months later, counts 766,757 transactions in 2025, +6.4%, a nearly identical figure. Italy's real estate market really is growing, regardless of how much artificial intelligence contributes to that, on which no verifiable sector-wide figure currently exists.

What this means for whoever decides

For someone running an agency and evaluating an artificial intelligence tool for sales, the useful question isn't whether striking numbers exist, but what to ask before signing: how many agencies and over what period each result the vendor shows was measured, and whether the same vendor states a different number elsewhere, on its own site or in other presentations, for the same benefit category. These are the same questions worth asking when evaluating any management software, not just an AI tool: I wrote about it in business management software. It's the same pattern I recently found in a case involving artificial intelligence and dental practices: a different sector, the same caution needed in front of a number published by whoever sells the tool.

If you're evaluating a tool like this for your agency and want a second opinion before deciding, get in touch: I'll help you read the numbers a vendor shows you, not pick the most talked-about tool.