This week wasn't a live deal-selling session - it was the official public launch of LiftGPT, Investorlift's new AI underwriting assistant, led by Robert Wensley after roughly 18 months of development and 6 months of internal testing.
What LiftGPT is: an AI assistant connected directly to Investorlift's own data (not a general chatbot layered on top of public scrapers). It pulls from nationwide MLS data (updated weekly), tax and deed records, mortgage data, HOA and permit history, crime and school scores, listing history and photos, and seven years of Investorlift's own marketplace activity - real buyer and seller behavior that shows up on the platform months before it ever reaches public records.
What it can do:
Generate ARV and comps as a range (not a single static number), with the underlying comps visible and adjustable - in internal testing against 1,000 past Investorlift deals, it landed within a median of $16K of actual resale price, versus $32K for Zillow's Zestimate and $100K+ for a competing AVM; typically within about $5K in major metros
Estimate repair costs using regional labor-cost data from the Bureau of Labor Statistics, adjustable by scope of work
Calculate max allowable offer for flips, with editable assumptions (financing cost, closing costs, holding period, etc.)
Run rental analysis - rent comps, cash-on-cash return, cap rate - with adjustable vacancy and operating expense assumptions
Pull people and ownership data: owner of record, mortgage balance and equity estimates, individual skip tracing, and reverse lookup by phone number or email, with litigator/DNC scrubbing built in
Surface local market trends: median sale price, days on market, sale-to-list ratio, and price-drop activity, with the ability to filter out distortions like new-construction sales skewing a neighborhood's numbers
Take natural-language follow-up questions - you can just tell it to reweight an assumption or ask "does this comp have a basement?" instead of rebuilding the analysis from scratch