Who was in the room

Over 40 members joined the first session, representing nearly every corner of commercial real estate: architects, brokers, attorneys, asset managers, project managers, marketing professionals, construction consultants, lenders, and finance professionals from across the U.S.

Tools members are currently using

• ChatGPT (most widely used) — drafting, research, brainstorming, summarization.
• Microsoft Copilot — valued for integration with Word, Excel, Outlook, and Teams.
• Claude — growing preference, particularly for long documents and reliability; one firm is actively implementing Claude Cowork.
• Google Gemini — Workspace-integrated workflows and renderings.
• Perplexity — research and citation-backed summaries.
• Proprietary internal LLMs — several firms have enterprise tools their IT teams approved.

Current use cases

Most members are using AI for email drafting and polishing; research, summarization, and brainstorming; marketing content and social media; writing and presentation support; and light data analysis and Excel assistance.

Emerging, more advanced use cases discussed: lease abstraction and clause extraction, construction drawing analysis, bid leveling, financial modeling, underwriting and data collection, Python/Revit scripting for architecture, and 1031 exchange research.

Key themes

Moving beyond basics

The clearest appetite in the room was for moving past polishing emails into more strategic, embedded applications — AI working inside workflows, not sitting alongside them.

Security & confidentiality

The most-cited concern across the group. ChatGPT was flagged as inappropriate for confidential client data. One firm is reportedly paying $500K/year for a secure enterprise solution — a cost gap that leaves most smaller firms without comparable protection. The need for affordable, secure alternatives came up repeatedly.

Output quality & reliability

Hallucinations generating false project leads, inconsistent quality, and the need for manual verification were common. Approaches members use to manage this:

• Running multiple models simultaneously to cross-check outputs (the "three to four horsemen" approach).
• Requiring full citations and bibliography verification.
• Using a second model to fact-check the first model's sources.
• Uploading bias documents to constrain and guide results.

Ethics & responsible use

Several members raised the need to define responsible AI practices before adoption accelerates further — environmental impact, workforce displacement, and data ethics among them.

Connections & action items

• Haley Mott (JLL) offered to share her law firm's formal AI platform evaluation with members navigating confidentiality decisions.
• Haley also volunteered to build a shared tool inventory spreadsheet — now part of the group's shared Google Sheet.
• Kenne Shepherd and Kristen Suzda connected to continue a conversation on responsible AI in architecture.
• Kerry Mason shared contact information for broader networking across the group.

What we decided

• Monthly sessions extended to one hour.
• CREW Network remains the primary communication platform (no Slack).
• Meetings recorded with Copilot; notes and summaries shared on CREW Biz.
• Member survey to follow Session 2 to shape the ongoing agenda.