Most of the AI conversation in private equity has been about sourcing, diligence, and value creation. But a quieter transformation is underway inside portfolio teams — closing the gap between when something goes wrong at a portfolio company and when the GP actually finds out. Stu is joined by Yann Magnan, CEO and Co-Founder of 73 Strings (AI-powered valuation and monitoring platform, backed by Goldman Sachs, Blackstone, Fidelity International, Golub, and Hamilton Lane), and Joe Zein, Co-Founder of Soal Labs (custom AI infrastructure for PE and private credit). One builds the platform. One builds the custom architecture that sits around it. Together they lay out why portfolio monitoring is uniquely hard to automate, and what operators should actually do next. You'll Learn: Why the real bottleneck in portfolio monitoring is infrastructure, not cadence — and why LP demand for weekly/daily reporting is about to force the issueWhat a governed data model actually requires (auditability, versioning, rules, entity resolution) and how it differs from a pile of Excel filesWhy private capital needs 100% accuracy, not 99.5%, and what that means for how LLMs get deployedWhen lighter, cheaper models beat frontier LLMs — and the case for combining ML, NLP and LLMs inside a single ingestion pipelineWhy "vibecoded" portfolio monitoring tools fail the moment they touch real LP reporting — and how to think about buy vs. buildThe entity-resolution problem (the same company named three different ways across 73 Strings, Salesforce, and your file system) and how to solve itWhat changes in 18 months: daily mark-to-market, AI-surfaced alerts in your inbox, and the prerequisite data foundation that makes it possibleChapters: 00:00 – Intro & the 60-day information gap 02:33 – Why monthly monitoring is brutal: data, process, bandwidth 06:10 – The real bottleneck at the GP: how reports actually get assembled 09:20 – Best and worst case timelines for a quarterly close 10:44 – Is the problem cadence, or infrastructure? 13:40 – Signals: what shows up before EBITDA moves 18:57 – New data sources, covenants, and the credit boom 21:31 – Why "vibecoded" monitoring tools fail 25:20 – The 100% accuracy problem in private capital 26:55 – What a robust data model actually looks like 29:45 – Moving from file systems to governed structured data 35:49 – Entity resolution across 73 Strings, Salesforce, and the file system 37:00 – Light vs. frontier LLMs: which do you actually need? 40:17 – Confidentiality, enterprise plans, and the open-source option 43:59 – What most people miss about AI and portfolio monitoring 46:29 – Portfolio monitoring in 18 months 49:35 – One thing to start doing right now 🎧 Listen on Spotify: https://open.spotify.com/show/3OlTNvh2FGlE4VJyCrMJVE 📺 Watch on YouTube: https://www.youtube.com/@JustCuriousAI 🔗 More Expert More interviews: https://checkpluris.com/expert-interviews Just Curious is the content and podcast brand of Pluris, helping leading operators and investors cut through a noisy AI market and find the right partners — from strategy and assessments to full-stack AI development. Our network of applied AI experts spans generative and agentic AI, workflow automation, and custom AI-powered software, helping teams turn curiosity into measurable value. Explore more interviews and connect with experts at https://www.checkpluris.com Subscribe to our newsletter at https://just-curious-ai.beehiiv.com