Agentic AI for Wall Street: Lessons from three YC companies

Executive overview

Brothers Arnie and Chas Englander built Model ML — an AI workspace for financial services — after selling two prior YC companies, Fancy and Fat Llama. The product replaces manual, repetitive analyst work by connecting a Word/PowerPoint/Excel-style interface to a firm's entire data ecosystem. Financial firms spent 2024 in testing mode; 2025 became the year of signed contracts and production use.

Firms no longer need to gather information manually — the system surfaces it before they ask.

What Model ML does

• Mirrors a human analyst's full digital access: files, emails, CRM, data vendors, filings, custom datasets.
• Builds documents (slides, earnings summaries, cover pages) that are 90–95% complete on release day.
• Replaces days of manual work per report; can pull from multiple sources for the same figure, increasing accuracy.
• Next shift: tasks that currently require a human to "click run" will happen fully autonomously overnight.

The market shift in 2025

• 2024 was pilots and proof of concepts; 2025 flipped to multi-year contracts.
• In one week they signed the same number of contracts as all of Q4 the prior year.
• ~10% of the largest private equity firms and investment banks globally are now customers; 80% of revenue is US-based.
• The buy decision sits at CEO or most senior executive level — not team or group level.
• Vision models and improved function calling made the biggest practical difference over the prior year.

How the product was discovered

• After selling their second company, they built an agentic deal-flow system for their own investing.
• An emailed opportunity triggered the system to pull LinkedIn profiles, Crunchbase data, review sites, and comparables — producing a one-pager 90% sourced from outside the original email.
• They realised they were better builders than investors and turned it into a product.

Selling to financial firms

• Demo live, on a laptop, with real data and real use cases — not a screen share.
• Trust is the key blocker: buyers risk their jobs on this decision.
• Face time and highly customised demos are how trust gets built.
• Offices in London, New York, Hong Kong, Singapore, and India to maintain proximity to customers and decision-makers.
• Many global technology decisions for financial firms are made from San Francisco, not New York or London.

Lessons from Fat Llama and Fancy

• Fat Llama (peer-to-peer rental marketplace with insurance): took three years to reach product-market fit; defined as unit economics working in a large enough market.
• Fancy (vertically integrated last-mile grocery): found PMF almost overnight — delivering beer and ice cream to students at corner-shop prices during COVID.
• Both exits looked like fairy tales from outside; internally each had constant near-disasters.
• Stripe shut down Fancy's payments for several days; they took orders by phone, cash, and PayPal.
• Doing the first 1,200–1,500 deliveries themselves meant they knew every route and heard every customer complaint in real time.

Perseverance and what it actually means

• Persevere when logic supports continuing — not out of stubbornness.
• Founders who apply unemotional first-principles reasoning and keep going tend to win.
• The worst-case outcome of starting a company is learning a huge amount in a short time.
• Jobs you pass up will still exist in a year; the opportunity to build early won't.

Hiring

• Enjoyment of working with the person now outweighs CV and prior employer.
• Ask what candidates have been building outside work — engineering or not.
• Hiring slowly and for cultural fit matters more as each person's output grows.
• People from finance backgrounds need to unlearn big-company thinking: stop making slide decks, move faster.
• "Build the plane on the runway" — ship fast, learn fast, stay lean.

Co-founder dynamics

• Venn diagram: Arnie owns engineering/product, Chas owns finance/commercial; overlap is customers and product.
• No filter between them — radical transparency prevents miscommunication.
• Founder fallout is the leading cause of startup death; trust and clear division of responsibilities prevent it.
• Interest alignment matters more than skill overlap: if both want to write production code, there's a problem.

Why do YC three times

• YC culture — reporting weekly or daily rather than monthly — never leaves a company.
• The peer network is irreplaceable, especially for founders outside San Francisco whose social circles don't share their work ethic.
• W24 was the first batch that was AI through-and-through; the Slack channel felt like daily scientific breakthroughs.

Where to base a startup

• San Francisco is the best environment to build: work ethic, peer density, and decision-maker access.
• London has strong, less competed-for engineering talent — useful for keeping hiring costs down.
• If staying in Europe, move to a tier-one city; don't settle for second-best talent.
• Europe is not one country: relocating within Europe can be as big a change as relocating to San Francisco.