Institutional AI, reimagined.
The investment function of an institution on one clean layer of data, with the tools that run on top of it. AI does the gathering, cleaning and drafting. People make the calls.
Everything the office receives lands in one place. It becomes one record. Every tool reads from that record, and people decide what to do with it.
The data is everywhere.
None of it agrees.
Too many sources.
Custodian files, manager statements, portals, legal documents and an inbox full of notices. Each in its own format.
Systems that don't match.
The custodian, the general ledger and the portfolio system all describe the same holding differently.
Rebuilt by hand.
Every month someone exports, cleans and re‑keys the same spreadsheet, and hopes nothing changed.
One layer underneath.
Every tool on top.
Inputs become one record. The record feeds every tool. The tools feed the decisions, and the decisions stay with people.
Ask once, answered from every document the office has ever read.
Custodian and accounting. What the statements say.
Not‑in‑bank assets. Commitments, direct holdings, side pockets.
Investment book of record. Both sources consolidated into one working view. What you actually hold today, and the layer portfolio management and risk run on.
General ledger and reporting. The reconciled record the books, the audit and tax reporting run on.
From a raw custody file
to a committee‑ready dashboard.
Five steps. The dashboard is the easy part. The clean data underneath is what makes every tool after it fast.
A scheduled download or the email the custodian already sends. MCP where a vendor offers it.
/custody/daily/Every field mapped to one master file. Formats cleaned, gaps flagged, every change versioned. Anything new waits for approval.
mapping v1.14147 columns become the 12 people actually use.
147 → 12Rebuilt every morning. Allocation against policy, plus look‑through for equity, fixed income and alternatives.
06:45 dailyCash forecasts, reconciliation, fee forecasts, drift alerts. All reading the same clean data.
same data- positions.csv✓
- transactions.csv✓
- cash_activity.csv✓
| Security | Class | Value | % |
|---|---|---|---|
| Lattice Core 500 Index ETF | Equity | $277,265,780 | 23.5% |
| Camden Holdings Cl B | Equity | $98,364,800 | 8.3% |
| Harlow Municipal Bond Fund | Fixed income | $22,450,887 | 1.9% |
| Cedar Ridge Partners IV | Alternatives | $38,620,000 | 3.3% |
| Meridian Govt Obligations MMF | Cash | $8,904,551 | 0.8% |
| 312 holdings · ties to custodian ✓ | $1.18B | ||
Twelve working tools.
Open any one.
Each one was built with AI against a real workflow, then rebuilt here with invented data. Open a tool for screenshots or the live sample.
The rest of the office,
on the same record.
Mapped, not built yet. Each one runs on the same layer as the tools above.
Human judgement
and decision making.
AI gathers, cleans, checks and drafts. It doesn't approve, commit or sign.
- New mappings wait for a person.Nothing enters the master file without approval.
- Every figure is traceable.Back to the document, the page and the file it came from.
- Recommendations are drafts.The investment committee makes the call.