Lots of Hype, Little Substance – where AI actually helps
The chatbot is the least interesting part of AI in real estate. Integrations of large language models (LLMs) should provide helpful tools connected to existing applications; they shouldn’t attempt to replace human judgement but rather assist a professional throughout their work.
The true work for AI is behind the scenes: it involves the tasks of keeping rent separate from operating costs during context retrieval, locating the relevant lease clause, and helping the user with various tools. AI and Agentic AI proves its value when it links a business question to the documents, data, and systems required to create a result that an asset manager can review and utilize.
The hype starts when a fluent conversation is mistaken for a finished business process. A chat window alone is not progress. What matters is whether the work produced by LLM leaves the portfolio team in a better position to act.
Connecting the question to the system in ICRS
LLMs do hallucinate and have no memory; therefore, the relevant context has to always be provided for AI. The ICRS Nuxxor AI’s RAG approach does context retrieval throughout the ICRS application, therefore always resulting in relevant and recent responses from AI. For rent roll questions, Metamagix’s AI RAG application can pull relevant documents and passages along with the references of the documents themselves for review. In addition to that, ICRS AI can also execute SQL queries and retrieve information from the databases by further augmenting its responses. The professional can always review the surrounding contracts or documents and make any relevant amendments to those documents as needed.
Trusting AI with access control or any security matters is not the best approach when integrating LLMs, as AI can easily be deceived. In ICRS Nuxxor AI, access control and security in general are handled by application logic, not delegated to the model. The application determines whether the user is authorized to perform a requested action or provide information.
- LLM (Large Language Model)
An artificial intelligence model trained on large amounts of text data to understand and generate human-like language. Examples include Microsoft Copilot, ChatGPT, and Claude. - RAG (Retrieval-Augmented Generation)
An AI architecture that combines a Large Language Model with information retrieval. It enhances responses by accessing relevant external data sources or documents before generating an answer. - SQL (Structured Query Language)
A standard language used to create, read, update, and manage data stored in relational databases. - MCP (Model Context Protocol)
An open protocol that enables AI models to securely connect to external tools, data sources, and applications, allowing them to access additional context and perform actions beyond their built-in knowledge.
metamagix’s DATransformer and Reporting custom tool
ICRS Nuxxor AI can interface with many tools; it can email reports, generate visualisations, create and edit Excel and Word files, and access outside data such as Eurostat. In addition to widely available tools, it also has many custom tools to interface with the ICRS application. For example, ICRS already has automated imports, data checks & reports. ICRS Nuxxor AI has its own integration to request any of the reports available in the ICRS and give it back to the user for review.
Another tool ICRS Nuxxor AI has access to is DATransformer, with the help of which it can suggest transformations, correct the suggested rules, and save them. The MCP server can then apply those rules on compatible files, without prompting the AI to infer the mapping again. The value is in both outputs: a usable file and a definition of the transform that explains how to transform the data.
In conclusion
The goal of connecting LLMs to the application should be to allow more natural interactions with the application and to automate routine processes in an agentic way. ICRS Nuxxor AI is able to pull a report, transform a file, or answer a rent roll question not because the model is smart, but because it is connected to the systems where the data actually lives. The integration is intelligent, not the conversation.
That connection to real systems is what separates a useful AI product from a demo. A stand-alone chatbot can summarize a lease. So can an analyst within 10 minutes. The difference is when the same question triggers a database lookup, retrieves the source document, and provides a formatted result the professional can act on without switching between three applications. That is where the time is saved and where AI deserves its place in daily workflow.
For real estate professionals evaluating AI tools, the question should never be “how smart is the chatbot?” It needs to be, “what systems does it actually get to, and does it make my work faster?” The technology is ripe for adding real value in commercial real estate, but it will only be when it is wired into the infrastructure behind the interface. Everything else is just conversation.