How Fifth Dimension Cuts Investment Memo Drafting from Weeks to 30 Minutes

Commercial real estate sits at the intersection of finance and asset management, where deals are defined by documents. A single acquisition can involve hundreds or thousands of contracts, leases, rent rolls, financial models, and legal filings. Before decisions get made, teams spend weeks reading, extracting, and synthesizing this information. Fifth Dimension was founded to eliminate that bottleneck using large language models and entered the market with a platform purpose-built for CRE workflows — acquisition intelligence, asset optimization, reporting, and communication.

The core technical challenge was reliability at scale. Commercial real estate workflows are not quick queries; they are long-running processes that can last between one and three hours, involving massive document datasets and requiring models to switch between tasks throughout the run. Fifth Dimension needed a foundation that could handle this without degrading, and it needed access to the best available models on the market without managing multiple infrastructure boundaries. The team chose Google Cloud and Vertex AI as its single secure environment.

Fifth Dimension built a multi-model orchestration strategy on Vertex AI that assigns tasks to the model best suited for each job. Gemini Flash and Pro models handle the large-context intake work — processing datasets that may span thousands of documents — using their one-million-token context window. Claude Sonnet and Opus models take over for high-precision tasks like calculating rental income adjustments, extracting structured data from complex documents, and generating Excel-based financial models. Claude Haiku handles cost-effective code execution on the fly. Gemini starts the pipeline with large-scale document grouping; Claude performs the subsequent high-performance extraction and reasoning. The switch happens automatically within the workflow. Running everything through Vertex AI means client data never leaves a single secure boundary.

The business impact is concrete. Investment memo drafting, which previously required days or weeks of analyst time, now takes 30 minutes. The platform has signed major deals with top-10 U.S. asset managers and grown to a 50/50 split between U.S. and international clients. Internally, every engineer at Fifth Dimension uses Claude Code for development, generating more than 300,000 lines of code per month with a 98% suggestion acceptance rate. Non-engineering staff, including account managers, use Claude Code connected to BigQuery to run real-time natural language queries against production data.

Fifth Dimension projects a 6x increase in inference spend in 2026 as its customer base scales. The company’s next phase involves internationalization and helping clients embed their own institutional knowledge directly into the platform — building what it calls each customer’s “secret sauce” into an AI-native operating system for real estate.