What Locari deliberately doesn’t do

Locari doesn’t rate people.

You set the criteria. Locari filters by your criteria – not by an algorithm’s guesswork.

There’s a small, technically crucial difference between “AI helps with letting” and “AI decides who moves in”. We chose the first in our architecture. This page shows what that means and how you recognise it in the product.

How Locari works Side-by-side comparison →

Anti-profiling in four numbers

Four markers – together, the architecture.

For us, legal compliance isn’t a clause in an appendix but the prerequisite without which the product wouldn’t have launched.

Art. 22

No automated decision

The General Data Protection Regulation (GDPR) prohibits automated individual decisions about people without human intervention. Locari is built so that this case never arises.

AI scores about people

No ranking, no matching score, no “top-match” label – Locari delivers no score.

100 %

You decide

Based on your criteria – you decide who moves in. Locari delivers preparation, not a verdict.

Database location

The database is in the EU. Which services work outside the EU is listed under sub-processors.

The four pillars

Anti-profiling, broken down into four structural guarantees.

No slogan. Four architectural decisions, made before the first line of code.

Following GDPR Art. 22

Invitations follow your criteria. Who gets the tenancy is decided by you alone – the person who lets.

Rule-based pre-selection

Pre-selection runs on hard facts you define yourself – number of people, income relative to rent, pets, move-in date. Traceable, changeable at any time, no hidden ranking.

Briefing, not a score

Before every viewing you get a briefing: what was said, which documents are on file, which of your criteria are met. No score, no star rating, no recommendation.

Database in the EU

The database is in the EU; the services outside it are disclosed (see “Privacy, concretely”). The three-phase model of the German Data Protection Conference (DSK) is reflected in the process.

The difference

Algorithmic scoring ranks applicants. We don’t.

Anyone rating applicants via AI automates the discrimination question. Click through our four steps – on the right you’ll see what happens under algorithmic scoring.

You set the criteria.

You determine the pre-selection criteria – number of people, income relative to the base rent, pets, employment. No learning model decides what counts: only your hard facts go in, traceable and changeable at any time. What isn’t on your list isn’t checked.

Locari

Algorithmic scoring

Concrete, no marketing fluff

What Locari does. And what Locari deliberately doesn’t.

On the left, what Locari takes on. On the right, what we deliberately leave out.

What Locari does

The work, without judgement.

Replying to applicant enquiries around the clock – in your tone, with your signature.
Collecting mandatory details (number of people, income, pets, move-in date) against your criteria.
Receiving required documents and checking them for completeness – no credit interpretation.
Filtering out unrelated messages, spam and hate, so you never have to see them.
Coordinating viewings, with reminders and reschedule links.
A briefing per appointment: what was discussed, which documents are on file, what’s open.
On critical actions, checking back with you via WhatsApp – special slot, short-notice reschedule, special condition.

What Locari doesn’t do

No algorithmic rating of people.

Assign applicants a score, a ranking or a “top-match” label.
Sort applicants by the likelihood of being a “good tenant”.
Translate personality, language style, photo impression or soft factors into a recommendation.
Algorithmically derive a creditworthiness verdict from income documents.
Use applicant data to train AI models.
Store applicant data longer than the letting requires.
Reuse or sell aggregated applicant statistics internally.

Legal context

Anti-discrimination law prohibits discrimination. An algorithm can’t override that.

An AI model that learns from past letting data learns its discriminatory patterns too – even when the protected characteristic isn’t in the data at all. Models find stand-ins such as postcode, name or language style. Harmless-looking input turns into a score that systematically disadvantages certain groups.

The only reliable way to comply with anti-discrimination law is to put no person-rating algorithm in between. Locari checks hard facts against your criteria – and nothing beyond. What isn’t rated can’t discriminate.

Privacy, concretely

GDPR as practice, not as a tick-box in the appendix.

Many tools work with long clauses and short practice. We do it the other way around.

1
The DSK three-phase model (V2.0, January 2026) in the process
The process runs in phases. Each phase only evaluates and requests what it actually needs – credit report and proof of income not before document review, not at first contact.
2
Database in the EU, services disclosed
The database is in the EU. For AI and delivery we also use services outside the EU, based on EU standard contractual clauses or an adequacy decision. Applicant communication (dialogue, reading documents) runs on Mistral (Paris), not OpenAI. The landlord assistant (e.g. briefings) additionally uses Claude (Anthropic, USA) and processes applicant master data in doing so. The full list is at locari.ai/legal/sub-prozessoren.
3
Data minimisation by default
We ask for as little as possible, as late as makes sense. No income documents before serious interest, no required documents before the viewing.
4
The landlord’s own signature in every conversation
Applicants write with you, not with an unfamiliar third party. A consistent presence and more privacy in one.
5
Data control stays with the people whose data it is
Applicants can have their data deleted at any time – we name the way right inside the conversation. When an applicant requests access under Art. 15 GDPR, Locari compiles the complete, readable report from the stored data and sends it within the statutory deadline. Access, correction, deletion: standard, not a special procedure.
6
Retention only as long as the letting needs it
Once the letting is over and statutory retention obligations expire, applicant data is deleted. The periods per data category are in the privacy policy.

What it’s about

What anti-profiling means in everyday letting.

Three principles that make the difference in every letting.

The core

No algorithm decides about people. You always know why someone is up for selection.

Your criteria

Pre-selection runs on your criteria – income, number of people, pets. Not on a learning model no one can explain.

Briefing, not a score

A briefing says more than a score. You decide based on information – not on a number.

Side-by-side comparison

Where we differ from AI letting tools.

The comparison fields as they come up in demos and sales conversations. Anyone choosing differently should do so on the basis of architecture – not marketing obfuscation.

Feature	Locari	Other AI letting tools
Implementing GDPR Art. 22	Anchored in the architecture – no automated applicant decision exists	Claimed, without naming the mechanisms
Pre-selection logic	Rule-based on landlord criteria (hard facts)	AI matching, scoring by a learning model, opaque ranking
Briefing per applicant	Preparation document, no score, no recommendation	Score, ranking or “top-match” label
Anti-profiling as a position	Explicitly stated as a principle	No explicit anchor, hidden behind “intelligent selection”
Final decision	Based on your criteria – you decide who moves in	AI-assisted with a recommendation
Data aggregation for training	Doesn’t happen	Often reserved as a possibility in the terms
Landlord signature in applicant dialogue	Per letting case, your signature	Platform signature, third party visible

The table is deliberately sober. The other providers have built serious products but made a different architectural decision – with real consequences for landlords and applicants.

Frequently asked

What we get asked most often in demos.

These questions come up in almost every first conversation.

Does “no scoring” also mean I get no comparative information at all?

No. For every application you get a structured briefing – which documents are on file, what was discussed, which of your criteria are met and which aren’t. What you don’t get is a ranking or a verdict. Information yes, the judgement is yours.

How can an AI product work without rating applicants?

Locari uses AI for the things AI is good at – understanding text, replying, extracting documents, coordinating appointments, summarising briefings. Not for the thing where AI is structurally problematic – judging a person.

How does anti-profiling relate to the economics of letting?

Behind AI applicant scoring lies the thesis that an algorithm can spot “better tenants” than a human. The empirical basis for that is thin – the models spot correlation patterns, often stand-ins for protected characteristics. A stable tenancy grows out of fitting criteria and a deliberate decision, not out of a score.

Who is liable if a letting process becomes problematic under anti-discrimination law?

The anti-discrimination risk sits with the landlord – with or without a tool. We don’t push you further into it, because we give no unexplainable sorting recommendations. If you have to justify a rejection, you do so on your criteria, defined upfront and in writing – a far better position than pointing to an algorithm that “ranked the applicant low”.

What happens to applicant data after the letting?

It’s deleted once the letting is over and the statutory retention periods have expired. The period depends on the data category – billing-relevant records stay longer, rejected applications shortest. The exact periods are in our privacy policy.

Anti-profiling as the foundation