Nori Agentic argues private assets broke the security master and only always-on agents can fix it. Amol Kapoor and Clifford Ressel made the case on AI Engineer, citing $240 billion in 2025 secondaries volume, up 50% from 2024, and a forecast of $32 trillion in alternative assets by 2030.
I buy the diagnosis more than the automation hype.
Kapoor grounded it in a Fortune 200 asset manager in New York with 50,000 assets and about 200 columns per row. The firm ran day-over-day change analysis on its security master and sent every changed cell for manual triage, at about an hour of document hunting per change. Kapoor said a Nori cloud agent now spins up on each change, pulls Box PDFs, Bloomberg data and a slice of email, assembles a cited explanation with a keep-or-discard recommendation, then pings the cell owner in Slack. The model runs locally on prem, and human review stays required, which cuts the work to about five minutes.
Ressel explained why the mapping is hard, and his public market contrast carries the point. Public equities have shared identifiers like Bloomberg FIGIs and LSE RICs for trading plus ISINs and CUSIPs for regulation, so even a GOOG versus GOOGL share class problem can be coded. A loan to a power plant firm or land by a hydro plant has no shared identifiers, no central vendor and no standard ontology, so Ressel said the asset is defined entirely by its agreements and amendments.
The fair pushback is that unstructured mapping still fails without ground truth, and finance will not accept fully autonomous writes to books and risk. Kapoor conceded that point in practice when he said to aim for 95% automation and keep a human in the loop for accountability. That concession helps him, because five minutes of review unlocks far more use cases than chasing 100% end-to-end autonomy.
Kapoor said the agents start as generic coding agents like Anthropic Claude Code, then get shaped with English markdown configs and the tools and credentials in the environment. Nori also frames that wider effort as Agentics, described in its writing as the study of how to use and reason about agents, per Agentics. If you sell to asset managers, build the audit trail first and the agent second.
