Introduction
In March 2024, Square Sense introduced its latest innovation: a virtual assistant for Real Estate Asset Management called AMAIA (Asset Management Artificial Intelligence Assistant). AMAIA offers prompt, accurate responses to queries about asset usage over any period.
However, AMAIA is more than just a question-and-answer system. It’s a sophisticated tool that interacts seamlessly with real-time asset data of any scale. This capability empowers clients to gain deeper insights into their assets and make informed decisions.
The introduction of AMAIA expands Square Sense’s service offerings, complementing existing solutions such as data centralization, visualization, and asset management support. By providing a conversational interface for data inquiries, AMAIA enhances client interactions and bridges the gap between data platforms and asset managers.
AMAIA brings together the precision of a data platform with the expertise of a data asset manager, at any time. This unique combination ensures that clients benefit from accurate insights without the delays typically associated with manual analysis and human resources availability. AMAIA’s user-friendly and interactive interface further enhances its appeal, making it an invaluable tool for asset managers navigating the complexities of real estate management.
Yet, this solution doesn’t aim to replace human expertise but rather serves as a complementary means to access information effortlessly and instantly, making the whole process more efficient and asset managers more independent. This also allows more room for Square Sense data asset managers to focus on what really matters: advising and accompanying clients in making decisions.
1. The Primary Challenge: Ensuring Calculation Consistency
During the development of AMAIA, Square Sense’s research team encountered numerous challenges. One of the most significant was ensuring that the Large Language Model (LLM) accurately understood the nuances of asset management to make precise calculations.
In asset management, key performance indicators (KPIs) often have specific definitions that differ from general contexts. For example, calculating average occupancy requires consideration of various factors such as weekends and holidays.
The term ‘average occupancy’ itself can be misleading and requires precise understanding by AMAIA. Due to the dynamic nature of real-time data, fluctuations are inevitable, such as a decrease in occupancy during lunchtime or late in the day. Calculating a true average may yield results different from what is expected. Asset managers typically focus on maximum daily occupancy rather than a simple average. This scenario illustrates just one of the many complexities involved, emphasizing the importance of expertise to avoid errors.
Failure to align our model with the expertise of asset managers may lead to, at best, a loss of client trust in our chatbot and, at worst, wrong decisions with damaging consequences for our clients.
To address this challenge, Square Sense needed a solution that would ensure calculation consistency and accuracy.
2. From SQL to Data-API
To tackle this challenge, Square Sense made a strategic shift from using traditional SQL tools to a specialized data API approach.
A significant number of data scientists and engineers regularly utilize SQL tools to navigate and query vast datasets, forming the backbone of most data analyses. Given this prevalent practice, it becomes logical for the LLM to replicate this task, translating and isolating user inquiries into SQL queries tailored to specific questions. Once identified, these queries are executed on the dataset, with the results then relayed back to the LLM, providing the necessary information to formulate a comprehensive response.
Despite the efficiency of this process, it faces a notable limitation—it can only perform generic computations. In other words, it falls short of delivering the specialized knowledge essential for Square Sense’s needs in asset management.
To confront this challenge, our full-stack team embarked on the initiative to develop a specialized data API for each KPI definition. Serving as an additional layer within the process, similar to an index, this approach ensures a direct connection to the data-API for any term related to a Square Sense-specific definition in asset management. By doing so, we can retrieve the requisite value based on specific parameters, such as a particular time period or spatial filters, enabling us to access unique insights tailored to our clients’ needs.
3. Benefits
I would like to emphasize the several advantages of employing the Data-API approach.
Firstly, it ensures heightened model accuracy, as critical asset values are computed by the reliable Data-API, rather than relying solely on the LLM. The LLM’s primary function is to interpret parameters and generate responses based on the data provided by the Data-API. This setup facilitates rapid execution, ensuring swift responses from the chatbot. Even complex inquiries, such as comparing weekly patterns across different periods, can be addressed within seconds.
Secondly, by consolidating the data set processing from the conversational interface, all Key Performance Indicator (KPI) values are exclusively obtained from the Data-API. Essentially, during a question-and-answer session, the LLM only possesses information relevant to the question, devoid of any knowledge regarding our clients’ dataset, including its schema and content. This methodology significantly reinforces the security of our client’s data. While the virtual assistant inherently operates as a private and secure tool, this adds a layer of protection concerning data privacy.
Conclusion
In conclusion, AMAIA represents a big step forward in real estate asset management, providing clients with a powerful tool for data-driven decision-making. By overcoming challenges in calculation consistency and harnessing the power of specialized data APIs, Square Sense has created a virtual assistant that combines precision with user-friendliness. As the real estate industry continues its digital transformation journey, AMAIA stands ready to support asset managers in navigating complex data landscapes and optimizing portfolio performance.
How about delving deeper into this topic?