Commercial Real Estate Deal Platform Duxre Launches AI Brief

Duxre, a commercial real estate deal platform, announced Tuesday that it has launched AI Brief, a software system that converts unstructured sales and leasing data into a readable large language model (LLM) format.

Founded in 2020, Santa Monica, Calif.-based Duxre is a subscription platform that aims to streamline data for CRE brokers to enhance transaction volume and client interaction.

AI Brief converts the unstructured files into Markdown files, which are more efficiently readable by LLMs such as ChatGPT and Claude, according to the company. The result, the company says, reduces file size from 50 megabytes down to between 30 and 40 kilobytes, speeds up processing, and minimizes errors.

“AI Brief specifically came from what we were seeing as we marketed assets for sale, and what our peers were realizing as they marketed assets for lease,” said Sean Fulp, founder and CEO of Duxre, as well as a vice chair at Colliers. (The two companies are not connected.) “Our deliverables, which are mainly in the form of a PDF, are being put into ChatGPT or Claude to analyze those deals. In the case of an offering memorandum — which can be 60 to 80 pages, and 15 to 20 megabytes in size — the first reader is no longer human, which these documents are created for. 

“There’s no sophisticated pipeline built for our industry for these documents to have an evaluation system in place to make sure that what’s pulled out of it is accurate. What we’ve been seeing is that the LLMs take these documents and are embellishing and hallucinating, causing issues with our process,” Fulp added. “These LLMs come back to tenants or buyers or brokers with a very convincing output, even if it’s wrong. You could say that the bid-ask spread just gets wider because of the disinformation.”

AI Brief works in three ways, Fulp explained: It begins with multi-source ingestion, where the brief is built from everything in the deal file, not just one document. That can include the offering memorandum or leasing brochure, the Excel financials, and the structured listing data the broker already entered such as the address, property details and further description. It synthesizes across all of those items and cross-references them.

The AI Brief uses code plus AI, with the language work done by a model. The math and structure are handled by deterministic code, so figures are calculated from the source rather than left to a model’s judgment. The brief never originates a number.

Also, an evaluation layer scores the output before it ships. It checks the extraction for accuracy and completeness. That’s the part nobody else in the chain does, Fulp said. When a buyer drops an offering memorandum into their own chatbot, that parse gets one unchecked pass and they never see what broke, he said.  AI Brief’s final version is a verified, multi-source brief, rather than a reformatted document.

As more CRE professionals are starting to use AI on a regular basis, they will begin to understand that Markdown files are the preferred format for these LLMs to ingest information, Fulp said.

“The difference is that a Markdown file is going to be 30 or 40 kilobytes versus 50 megabytes,” he explained. “So the file size is about 95 percent smaller than what an offer on a leasing brochure would end up being. A smaller file size means that it can load into ChatGPT, Claude or Gemini much faster. It costs less because the cost of ingesting a PDF can be 15,000 to 20,000 tokens, whereas ingesting a Markdown file could be 1,000 or 2,000.”

In addition, Marketplace Duxre, which sits on the Duxre website, offers a free subscription to create sales or leasing listings, he said. Brokers can publish a listing and generate a brief for that listing.

Philip Russo can be reached at prusso@commercialobserver.com.