Tenant screening determines whether a rental applicant is likely to pay rent reliably, maintain the property, and comply with lease terms. Poor screening leads to rental arrears, property damage, and costly eviction proceedings. Traditional screening relies on credit scores and reference checks, which are slow and provide limited predictive power. AI screening integrates multiple data sources for faster, more accurate risk assessment.

The Problem

Property managers face an asymmetric risk: a good tenant generates stable income for years, while a problematic tenant can cost 10,000-30,000 EUR in lost rent, property damage, legal fees, and void periods. Traditional screening uses credit scores as the primary filter, but credit scores were designed for lending risk, not rental risk. A tenant with a low credit score due to medical debt may be an excellent renter; a tenant with a good credit score may have a history of lease violations at previous properties.

Reference checks from previous landlords are unreliable: current landlords may provide positive references to facilitate a problematic tenant’s departure. The screening process is also slow - manual verification of employment, income, and references adds 3-7 days to the application process, during which desirable properties may lose qualified applicants.

AI Approach

Document verification - Textract processes application documents: pay stubs, bank statements, employment letters, and identification documents. The system extracts key data (income amounts, employer, account balances) and flags potential document manipulation (inconsistent fonts, altered figures, mismatched formatting).

Risk prediction modeling - SageMaker models predict tenant risk (probability of default, late payment, or lease violation) from a combination of features: credit data, income-to-rent ratio, employment stability, rental history, application behavior patterns, and property-specific factors. The model is trained on historical tenant outcomes from the property management portfolio.

Income and affordability analysis - Beyond simple income-to-rent ratios, the model analyzes income stability (consistent vs. variable), source reliability (employment vs. self-employment vs. government benefits), and overall financial health indicators from bank statements. The affordability assessment accounts for the applicant’s total obligations, not just the proposed rent.

Rental history scoring - Where rental history data is available, the model incorporates previous tenancy duration, payment patterns, and outcome (renewal, voluntary departure, eviction). Bedrock processes reference responses and landlord communications to extract sentiment and factual claims.

Architecture

Applications are submitted through the property management platform and stored in S3. Textract processes uploaded documents. Lambda functions orchestrate the screening pipeline: document verification, data extraction, credit bureau API calls, and risk model scoring on SageMaker. Results are compiled into a screening report generated by Bedrock and stored in DynamoDB. The report is surfaced to the property manager through the management platform with a recommendation and confidence level.

Key Considerations

Fair housing compliance - Tenant screening must comply with anti-discrimination laws. The model must not use protected characteristics (race, religion, national origin, family status, disability) as features, and outcomes must be regularly audited for disparate impact across protected groups.

Transparency and adverse action - Applicants denied based on screening results are entitled to know the reasons. The model must provide explainable reasons for adverse decisions, not just a risk score.

Data protection - Tenant applications contain sensitive personal data. Processing must comply with GDPR, including data minimization, purpose limitation, and applicant rights to access and correction. Automated decision-making under Article 22 requires appropriate safeguards.

Cross-referencing - Tenant screening shares risk assessment patterns with credit scoring in finance, underwriting automation in insurance, and the broader category of customer onboarding risk evaluation.

Next Steps

Analyze historical tenant outcome data to build the predictive model. Validate the model for both predictive accuracy and fair housing compliance using disparate impact analysis. Deploy in advisory mode alongside existing screening processes for a 6-month pilot, comparing model predictions against actual tenant outcomes. Transition to primary screening tool once accuracy and compliance are validated.