Old Capital Real Estate Investing Podcast with Michael Becker & Paul Peebles · Michael Becker & Paul Peebles

EPS 341 - "AI Meets Real Estate: Transforming Underwriting and Investment Decisions"

·48 min·2 clips
If AI handles the underwriting analysis, Eddie argues, what matters most becomes "who are you as a person? Are you trustworthy?"
1. Old Capital Real Estate Investing Podcast, Episode 341, is a videocast on how artificial intelligence, which Paul Peebles calls still in its "infancy," is being applied to commercial real estate underwriting. 2. Paul Peebles, Old Capital's National Underwriter, hosts in place of Michael Becker, who is on a due diligence assignment, and he interviews Eddie, an investor working in Silicon Valley big tech, and Tony Lin of Link Multifamily, whom Paul heard speak in Silicon Valley. 3. The core thesis is that AI lets investors gather more information and analyze more variables for lower-risk decisions, while Paul defines underwriting as risk assessment built on "prudent risk" rather than gut and experience alone. 4. Eddie traces the modern era to the 2017 paper "Attention is All You Need" and "next token prediction," describes 2018-2020 chatbots as autocomplete, and says tools like deep research now search economic and jobs reports. 5. Eddie separates hardware makers such as NVIDIA, TSMC, Samsung, Apple and Google from software like Claude, and mentions personal agents such as the Claude bot, the Perplexity computer and Mac minis selling off shelves for tasks like email triage and scanning broker sites for new OMs. 6. Tony Lin, a software veteran of about 23 years who retired roughly four years ago, sorts AI into three levels (general chatbots, generative AI and agentic AI) and says small and medium businesses sit in "a vast emptiness" between casual ChatGPT users and big-tech builders. 7. In his rent roll demo, Tony uploads a roughly 300-unit file to Claude and gets a 12% move-in rate over three months, a 24-cent-per-square-foot premium for renovated A1R units, 28% loss to lease on three-bedroom C1 units, 47 leases expiring in 90 days, and nine expired leases likely month-to-month. 8. Tony also has Claude draft renewal letters conditionally, flat rent if many tenants are leaving and $25 increases if few are, and uses NotebookLM, which he calls a RAG, to ingest an OM, T-12, rent roll and loan quotes and generate a briefing document and an 11-minute AI podcast. 9. Eddie says hallucinations were "a big problem" five to seven years ago but are far less of a concern in NotebookLM, which uses a vector database and answers only from uploaded files, and Tony adds that it flagged an $88 scheduled charge and duplicate balances on a new deal's rent roll. 10. Real estate investors and operators curious about practical AI underwriting workflows will get the most from this.
Listen to the show on