Price:

AI & TECH

AI agents automate $100 trillion fund infrastructure

Tuesday, July 7, 2026 · from 1 podcast
  • AI agents with memory replace the manual accounting and data silos that currently manage global wealth.
  • Long-horizon models compress multi-year fund migrations into days, unlocking automation for complex fund structures.
  • Pricing shifts from SaaS seats to basis points on assets, aligning vendor incentives with client outcomes.

Chris Halanish calls the $100 trillion asset management industry a stack of “human duct tape.” The CEO of Hanover Park argues that institutional data is held hostage by service providers using QuickBooks and Excel, creating a fragile system where a fund accountant's departure can wipe out its operational memory. He proposes replacing it with AI agents equipped with long-term memory to learn each fund’s specific legal quirks and accounting rules.

This is a move beyond selling software. Halanish is building an AI-native services company that owns the outcome. The goal is to automate the unsexy core systems of record - capital calls, profit-and-loss allocations - rather than just providing a tool for human operators.

The technical barrier was ontology. Mapping the complex web of relationships between legal entities, limited partnership agreements, and portfolio companies made migrating a fund like Blackstone a two-year project. Halanish claims this prevented modernization until roughly six months ago.

Hanover Park uses long-horizon agents to clean financial data at scale, ingesting hundreds of thousands of legacy documents to extract governing rules. In one case, his team migrated a fund with 20 entities in six days - a task his engineers considered impossible a year prior. The shift became viable with models like OpenAI’s Opus series, which handle massive context windows and maintain accuracy across decades.

"It's human duct tape... it's QuickBooks, it's Bill.com, it's Excel."

- Chris Halanish, This Week in Startups

This automation enables a pricing revolution. The era of charging for software seats is fading. While Figma’s Dylan Field built trust through seat transparency in 2020, new AI-native firms like Hanover Park are charging basis points on assets under management.

Outcome-based pricing aligns incentives. Legacy SaaS vendors profited from “shelfware” - seats paid for but unused. Charging based on AUM or value created makes the AI provider a growth partner. Hanover Park grew from $1 billion to $20 billion in assets under management over 15 months with a 50-person team, focusing on making CFOs “raving fans” rather than sharing aggregated data.

The model abstracts the labor. The AI agent becomes the institutional memory, eliminating the volatility of human turnover. For a system managing a century’s wealth, continuity is now automated.

Source Intelligence

- Deep dive into what was said in the episodes

$100T is managed by “human duct tape” | E2308Jul 6

  • Chris Halanish argues that $100 trillion in global assets is managed by a 'human duct tape' stack where legacy service providers using QuickBooks, Bill.com, and Excel withhold data from funds.
  • Hanover Park built an AI-native fund administration service, replacing manual accounting with AI agents that learn fund-specific rules to automate financial reporting, capital calls, and data cleaning.
  • Complex fund structures, like Blackstone’s hundreds of interlinked entities with unique profit allocations, make ledger construction difficult. Traditional migrations could take 24 months, but Hanover Park’s agents completed one for a venture fund in six days.
  • Chris Halanish states the shift to AI-powered data cleaning and ontology mapping became viable only 3-6 months ago, enabled by OpenAI's Opus model series, moving the industry closer to a 'one-click migration' future.
  • Hanover Park charges funds a transparent, all-inclusive fee based on AUM, contrasting with legacy administrators that levy hidden per-transaction and hourly rates for services like capital calls.
  • The company grew from $1B to $20B in AUM over 15 months, scaling its team to 50 people. Halanish focuses on making CFOs 'raving fans' rather than sharing aggregated fund data externally.
  • Dylan Field explained Figma’s early bottom-up growth strategy, where designers adopted the tool individually and brought it into organizations, easing enterprise adoption and reducing sales friction.
  • Jason Calacanis described Mahalo’s model of manually curated search pages, which reached $10M annual revenue before Google's Panda update wiped 80-90% of its traffic and abstracted its content into search result one-boxes.
  • In March 2020, Jason Calacanis predicted a swift COVID recovery, suggesting businesses would reopen by April 15th, while Dylan Field anticipated longer lockdowns and potential enforcement measures.