Innovation4 min read

AI in property management: practical uses and safeguards

What AI really does in an inspection report or in portfolio monitoring, where to start, what to check and which mistakes to avoid.

The homePad team·5 January 2026

In the tools property professionals use, artificial intelligence is used today to read a photo, transcribe a voice, draft a text and suggest a cost estimate. It replaces neither the visit nor the judgement of the person carrying it out. It prepares a draft that the professional corrects and approves. Used well, it cuts data entry and makes reports more consistent from one member of staff to the next. Used badly, it produces findings that nobody has really reread. Here is how to introduce it, step by step, in an agency, a property management firm or a real estate investment company.

What AI does today, and its limits

In the field, the common uses fall into four families:

• Recognising what is in a photo: a radiator, a window, a crack, a stain
• Transcribing what you dictate, so that the observation is written while you look at the room
• Drafting a first version: a description, a report summary, a list of works
• Estimating an order of magnitude of the cost from the defects recorded

In every case, the result is a suggestion. AI knows neither the history of the home, nor the lease, nor what was agreed with the occupant. It can take a shadow for a stain, miss an element left out of frame or misunderstand a word dictated in a noisy stairwell. Nor does it settle questions of law. In France, for example, telling damage apart from wear and tear, that is, the wear due to time and normal use, remains a matter for the professional's judgement. The rules differ in Switzerland, Belgium and Luxembourg and are to be checked locally.

Start with a single use

Rolling out every feature at once stops you from knowing which one is actually useful. First choose your teams' most repetitive task, the one whose result can be checked at a glance, then proceed in this order.

Test each use on a few representative properties, with two or three volunteers from your staff, before extending it.

Organise human approval

The rule is simple: nothing goes out to a tenant, an owner or a contractor without having been reread by the person who carried out the visit. Spell out what this review must cover:

• the list of elements in each room, with nothing missing and nothing duplicated
• the condition given to each element and the vocabulary used
• the link between each photo and the observation it illustrates
• the quantities and surface areas carried over into a works report
• any sentence that assigns a responsibility to the occupant

Decide who signs, too. The report commits the company, not the tool. A finding generated and then signed without being reread is hard to defend if it is disputed.

Check where the data goes

The photos of an occupied home, the voice of a member of staff and the name of a tenant are personal data. In the European Union, the GDPR applies to their processing by an AI as it does to any other processing. Before choosing a tool, put four questions to its vendor:

• where the data is processed, on the device or on a server, and in which country
• whether it is used to train models, and how to object to that
• how long it is kept
• which data processing agreement governs the service

Then enter this processing in your record of processing activities and update the information given to occupants.

The most common mistakes

Confusing an estimate with a quote. A cost suggested by AI serves to prepare a decision. It binds no contractor and does not by itself justify a deduction from a deposit.

Neglecting how photos are taken. A photo that is dark, blurred or too wide gives poor detection.

Letting the vocabulary drift. If everyone approves different wordings, reports stop being comparable from one visit to the next, and portfolio monitoring suffers: data structured in the same way can be compared, free text cannot.

Measuring nothing. Record your own figures before and after: time spent per report, number of corrections at the review stage, disputes received. They are worth more than a vendor's promises.

Aiming too far, too soon. Predictive maintenance or the digital twin require a history of reliable data. Start by producing that data.

What homePad offers

The homePad mobile app includes AI features built on one principle: AI suggests, the professional approves. It recognises elements and defects from photos, writes the observations you dictate, carries out voice commands, prepares a works report, summarises reports and suggests an estimate of the cost of the works. Every suggestion remains to be reread and approved by the professional.

• artificial intelligence
• inspection report
• portfolio management
• PropTech