Our client, an administrative property manager, faced a challenge with their current operations.

Their team of 15 employees dedicated a significant portion of their time to administrative tasks and responding to tenant, vendor, and owner inquiries.

Approximately 80-90% of their workload revolved around answering questions.

The client was looking for a solution to automate parts of their communication processes, beginning with the development of a Q&A bot capable of addressing a wide range of tenant inquiries.

To address the client's challenge, we implemented an automated communication system utilizing a large language model specifically designed to handle tenant-specific questions. The solution involved the following steps:

1. Data Preparation: Initially, we prepared the available unstructured data for each tenant in compliance with data privacy regulations. The data was segmented and stored in a vector database, enabling us to do a semantic search over a large body of data.

1. Data Retrieval Logic: We developed and implemented a robust data retrieval logic that fed into the large language model. This allowed the system to retrieve relevant information and provide accurate responses to tenant queries.
2. User Interface and Backend Integration: To facilitate seamless interaction, we designed a user interface specifically tailored for tenants. Additionally, we developed a simple backend system to enable easy tenant data management for the property manager.
3. Request escalation and Summary Generation: In situations where the bot could not address a query, the system could escalate the question to the property manager's ticket system. To streamline the manual processing, the system generated a summary of the inquiry for the property manager's convenience.
4. Authentication and Data Security: To ensure secure access, we implemented an email login mechanism, which was separately activated and linked to the corresponding tenant's dataset. This authentication process guaranteed the privacy of the data.

To further improve data privacy challenges, we used a technique called Named Entity Recognition (NER) to remove personal information from data whenever necessary. That further improves GDPR compliance.

Demo

‍

Through a pilot phase and subsequent refinement, the automated communication system yielded significant results for our client.

Initially, the bot successfully answered approximately 50% of tenant questions without requiring the intervention of the property manager.

However, as the system accumulated more data and knowledge, this success rate increased to 70%, enabling the bot to autonomously handle a majority of tenant inquiries.