E95: How Purelend Uses AI to Turbocharge Mortgage Apps w Lucas Scheer

AI in mortgages that actually saves time, not hype. On this episode of Mortgage Tech Talks I sit down with Lucas Scheer, co-founder of Purelend. We dig into what AI can do today for brokers, why trust still wins in real estate and lending, and how document collection, usable income, and down payment verification can be automated without losing the human touch. What you’ll learn The real advantage for brokers who use AI vs those stuck in manual review How Purelend structures down payment checks, stated income analysis, and doc organisation for lenders and brokers Where full automation fails and why a copilot model makes more sense Accuracy, hallucinations, and how to combine human review with AI to raise overall quality Practical privacy and workflow considerations when rolling AI into your shop Chapters 00:00 AI will not replace you. People using AI will 00:26 Set up for the episode 00:40 Intro and guest welcome 01:01 What Purelend does for brokers and lenders 02:24 Lucas’ background, NEO Financial, and the path to Purelend 03:10 The 60,000 km motorcycle detour and lessons 04:10 Building applied AI before ChatGPT 05:13 Pre-2020 AI adoption vs the post-ChatGPT shift 06:19 Focusing on outcomes over buzzwords 07:33 Media hype, fear, and what actually matters to a brokerage 08:28 Useful analogies for explaining AI to clients 09:03 Common fears brokers raise 10:11 Two brokers, two results. Why leverage wins 11:26 Excel analogy and why AI is the new table stakes 12:07 Privacy, data, and realistic risk 12:46 Will AI do prospecting and client calls 13:39 Why trust and emotion keep humans in the loop 15:14 Where automation helps and where it breaks 16:20 Purelend elevator pitch 16:46 Duplicate effort across brokers and lenders 17:46 Speed, cost basis, and winning deals 18:10 Why this only became feasible recently 19:18 Feature set: down payment, usable income, organising docs 20:02 Stated income launch and common gotchas 21:06 Borrower alerts without adding friction 21:39 Full auto vs copilot. Drawing the line 24:12 Specific knowledge that AI cannot replace 25:38 Borrower portal vision and UX tradeoffs 30:44 Accuracy, hallucinations, and safeguards 32:27 Probabilistic vs deterministic and hitting 92 to 96 percent 35:04 Benchmarking against human accuracy 38:50 Supervisors, different error patterns, and combined accuracy 40:38 Where to learn more 40:55 Close