Alex Lopez

Custom GPT / Claude Project

A customized AI assistant built on a foundational model such as OpenAI’s GPT or Anthropic’s Claude, configured with specific instructions, knowledge files, and behavioral guidelines for a particular use case. In commercial real estate, professionals use custom GPTs and Claude Projects to create purpose-built tools for tasks such as lease abstraction, offering memorandum drafting, market report summarization, and underwriting Q&A. These tools require no coding and can be deployed in minutes, making them among the most accessible entry points for CRE AI adoption.

Putting Custom GPT / Claude Project in Context

An acquisitions team builds a Claude Project configured with their firm’s underwriting criteria, standard assumption ranges, and a library of past deal memos, so that when an analyst pastes in a new offering memorandum the tool immediately flags assumptions that fall outside the firm’s parameters, drafts a preliminary investment summary in the firm’s format, and surfaces comparable deals from prior underwriting without requiring any prompt engineering from the analyst each time.

Frequently Asked Questions about Custom GPT / Claude Project

Using a general-purpose AI tool directly requires the user to provide context and instructions in every conversation, which creates inconsistency and puts the burden of prompt quality on each individual analyst. A Custom GPT or Claude Project bakes those instructions, context documents, and behavioral guidelines in permanently, so every team member interacts with a tool that already understands the firm’s terminology, preferred output formats, and task-specific constraints without any setup required on their end.

The most useful uploads are documents the tool will need to reference consistently, such as a firm’s underwriting criteria, standard lease abstract templates, a glossary of internal terminology, sample offering memoranda in the firm’s preferred format, or market reports covering the firm’s target geographies. The goal is to give the tool enough context that its outputs reflect the firm’s actual standards rather than generic industry defaults, reducing the editing burden on the analyst reviewing the output.

Custom GPTs are built on OpenAI’s models and configured through the ChatGPT platform, while Claude Projects are built on Anthropic’s Claude models and configured through the Claude interface. Both allow users to provide system instructions and upload reference documents, but they differ in how they handle long documents, maintain context across conversations, and respond to nuanced analytical prompts. The practical choice often comes down to which underlying model performs better on the firm’s specific tasks, which is best evaluated by testing both against a representative set of real documents rather than relying on general benchmarks.

These tools operate within the context window of the underlying model, meaning very large document sets or long conversation histories can exceed what the tool can process in a single session. They also do not connect to live data sources by default, so a tool configured with a market report from six months ago will not reflect current conditions unless the file is updated manually. For workflows requiring real-time data, system integrations, or processing at high volume, a custom-built API-based solution is generally more appropriate than a no-code configured assistant.

The most reliable approach is to time the manual process for a representative task before deployment and compare it to the total time required with the tool, including the time spent reviewing and correcting the output. A tool that produces a first draft requiring significant rework may not be saving as much time as it appears. Tracking the edit rate and the types of corrections made over the first several weeks of use also reveals whether the system prompt needs refinement, which is often the faster fix when output quality falls short of expectations.

Click here to get this CRE Glossary in an eBook (PDF) format.
by