How AI redefines contract work in property management

Starting point

Large property management companies often manage hundreds or even thousands of complex lease agreements. These agreements, including annexes and amendments, form the basis for property management work and communication with tenants.

Yet these documents are often available only on paper or as unstructured scans. This makes efficient use much harder, leading to substantial time expenditure, low tenant satisfaction, errors and increased operating costs.

A property management company operating across Europe faced exactly this challenge. Its central question was:

Can the AI solutions available today significantly improve the efficiency of our property managers, without a structured data foundation and without compromising privacy or security?

From the challenge to a tailored solution

The market offered no suitable standard solution to efficiently digitise and make use of the many individual, complex lease agreements, often available only as scans.

evoliq's role was to make a tailored solution possible and implement it as a strategic and technical partner. From the initial idea to a production application, we supported the company through all the decisive steps:

1. Analysis of processes, available data and organisational conditions
2. Identification of specific, immediately achievable use cases with high value
3. Development of a functional prototype (MVP)
4. Agile development following the “Build – Measure – Learn” principle

Implementation: focused on users and partnership

Close collaboration with the business team and the company's openness to innovative approaches were key success factors.

Future users, particularly property managers, were actively involved from the start. Prototype testing phases, short feedback cycles and a consistent focus on real everyday work produced a solution with broad acceptance and high value in production.

Key functions of AI-supported contract management

• Hierarchical data organisation: Management by administrative units, buildings and tenants, including straightforward navigation and search.
• Generative AI for automated interactions: Employees can work with contracts through chat, request relevant information and generate automated summaries.
• Optimised OCR technology: Documents are automatically digitised, made machine-readable and provided to subsequent workflows through a database.
• Mobile access from any location: A responsive, user-friendly interface supports the mobile ways of working of property managers.
• Private cloud infrastructure: Sensitive contract data remains protected within the customer's own private cloud in compliance with privacy requirements.

Results and benefits at a glance

• Up to 70% less effort in key work processes and routine tasks, particularly information searches and contract analysis.
• Contracts are actively used: Contract information is available on mobile devices at any time, whether at an on-site appointment or on the phone, and immediately usable in the current context.
• Better decisions, higher quality: All contract data is accessible at any time, improving tenant communication, increasing service quality and reducing errors.
• More focus, less overload in everyday work: Employees experience tangible relief, reducing chronic overload and increasing satisfaction.

Transferability to other industries

The evoliq solution was developed specifically for real estate, but its underlying principle applies across industries: wherever complex, often unstructured contracts or documents need to be made efficiently usable.

Successful tests with industry-specific document types show strong potential for further applications:

Conclusion: evoliq's approach can be applied across industries wherever document-based processes are held back by manual work, lengthy searches or a lack of transparency.

Privacy and integration with existing infrastructure

A central project requirement was full integration into the customer's existing Microsoft Azure-based private cloud infrastructure.

This approach met demanding privacy and compliance requirements, making it substantially easier to implement contract management supported by generative AI models (LLMs).

It also enabled a high degree of flexibility in adapting to internal processes and an optimal mobile user experience.

Technology: modular, scalable and controlled

The technological foundation combines modern web technologies with tailored AI components developed specifically for contract management in property management.

Technology stack

• Frontend: A progressive web application based on Nuxt (Vue.js) makes further development straightforward and provides a performant, mobile-optimised interface for all devices.
• Backend and storage: All server-side components and APIs use Python and FastAPI. MongoDB provides data and object storage, while Elasticsearch's vector storage functions form the foundation of RAG workflows.
• Generative AI: OpenAI chat and embedding models from Azure OpenAI Services support RAG workflows and AI agents that meet compliance and privacy requirements, with demanding expectations for LLM precision and performance. The models were integrated using the LangChain framework.
• OCR component: Tesseract OCR with a specially trained model provides precise recognition of legal contract documents originating from scanned paper agreements.

Reliable AI agents through guardrails

Reliable AI answers were a priority from the start. Large language models used generically without appropriate rules can hallucinate and provide speculative answers. The AI agents here are secured through a multi-stage guardrail concept:

• Answers grounded in data: Answers that are not based on the extracted data of the relevant contracts are not permitted. Interpretation, invented additions and model-generated assumptions are prevented.
• Explicit answer logic: If information is missing or ambiguous, agents state this transparently: “This information is not contained in the contract.” This avoids an illusion of knowledge.
• Auditability: Every answer is traceable and based on clearly documented sources within the platform. Original passages from source documents are included as references in the answers generated for users.
• Monitoring and improvement: Regular review of generated answers and continuous collection of user feedback improve the system and allow errors in the underlying data to be corrected.

This helps prevent incorrect AI answers from leading to wrong decisions or misunderstandings of information relevant to liability.