Built, shipped, and audited.
Each of these was built inside a regulated financial institution, shipped into a live decision workflow, and measured afterward. Client names withheld by professional convention; situations, architectures, and results exactly as delivered.
An underwriting engine a regulator could audit
Unstructured medical and application files to a bound decision, inside a HIPAA- and SEC-regulated environment.
99.9% accuracy · 33% faster → Agentic workflow · Private creditA 26-day monitoring cycle, compressed to three
A month-long human process re-cut as an agent workflow with human approval preserved at the judgment points.
26→3 days · 8× frequency → Data platform · Asset managementThe data model under a 1.5-billion-record moat
Entity resolution and platform modernization that turned overnight batches into same-morning answers.
−97% runtime · −35% cost → Predictive models · Insurance$754M of underwriting & reserve accuracy
A global analytics function built from a standing start — 160+ people across three continents.
$754M gain · 160+ people → AI leadership · Asset management$2.2B attributed to a governed AI portfolio
AI run as a capital-allocation problem: business cases, kill criteria, and board-grade attribution.
$2.2B+ profits · 14% IRR → Your firmThe next one starts with a conversation
None of these started as a demo. They started with a process that was too slow, too manual, or too hard to defend.
Bring yours →Original analysis, from public data.
Teardowns of real filings rather than commentary. Every number is reproducible from public sources — the method is stated so you can check it.
BDC filing velocity
Thirty BDCs, every periodic filing timed against its statutory deadline. The median fund files 38 days into a 40-day window — a two-day buffer on a process that can’t be compressed by working harder.
30 funds · median 38/40 daysAdviser entity sprawl
Twenty-six private capital managers mapped through their registered entities. Median: three to four. Top quartile: ten or more — each a separate reporting surface and a compounding reconciliation tax.
26 managers · median 3–4 entitiesThe State of AI Readiness in Private Credit
A benchmark study from structured interviews with 25 private credit COOs and CFOs. Participants get the findings first, plus their own firm’s position against the benchmark.
Participate