Before
Quarterly re-keying
The whole book is re-typed each quarter. One exposure question once took two analysts up to two weeks to answer.
Private equity and real estate in a single store, every figure traceable to the document it came from and the date it was true, with statements and capital calls that read themselves. Then you just ask, in plain English, right in Slack.
The investment book is rebuilt by hand every quarter from statements, portal exports, bank downloads, and email. It works right up until it doesn't: the one person who owns a sheet is away during a close, a number can't be traced to a source, or "the portfolio as of March 31" simply can't be answered. The platform replaces the spreadsheet with a governed store, and keeps the humans in the loop where judgment belongs.
The whole book is re-typed each quarter. One exposure question once took two analysts up to two weeks to answer.
A single person owns and interprets each sheet. The failure showed up live, during a close, while that person was away.
Numbers can't be tied to a source or an as-of date; copies drift, so "yesterday's sheet" and today's disagree.
NAVs mix quarter-ends, so "as of a date" and "what changed this quarter" are both unanswerable.
One fund is many rows across entities; PE and CRE live apart; total look-through exposure stays opaque.
The same entity has four names across accounting, the portal, the sheets, and conversation. Nothing joins cleanly.
One time-series store over the whole book, with fund-level consolidation from day one and total look-through exposure in a single query.
"Total exposure to senior living, look-through included." Answers come back cited, in Slack or on a dashboard.
Statements, capital calls, and distributions are extracted automatically across a proven set of document types, then a person confirms before anything lands.
Each number carries the document it came from, its data class, and its as-of date. Reporting stops being an act of faith.
The live deal pipeline runs in Linear, sourcing through close; agents draft and label, and only a human moves a deal's stage.
Extracted capital calls and distributions post with their source links, and become deadlines and briefing items downstream.
A weekly report surfaces what arrived, what's gone stale, and what's missing, so gaps are caught before someone needs the number.
Coverage and loan-to-value thresholds and upcoming maturities are watched for you, and surfaced before they become a scramble.
Role-based access from day one, figures classified and never over-shared, and a full audit trail on every change. Governance is enforced, not promised.
Statements, capital calls, and distributions land from sanctioned senders. Extraction agents pull the figures; anything that fails validation routes to review with a best-effort draft attached. Nothing is invented, and no one re-types.
Extracted figures surface in a Slack review channel as claim-and-resolve cards. A human is the verification layer: confirm, correct, or reject. Duplicate copies of the same document from different inboxes are absorbed automatically.
Confirmed figures enter the time-series store with their document, class, and as-of date attached, reconciling to the accounting system of record rather than replacing it.
Query in plain English from Slack or a dashboard. Every figure comes back with its source and as-of date, and anything above your clearance returns a dashboard link, never the number.
These are the real prompts, stated as the person would type them. If the platform answers these, it has done its job.
Role tiers from day one, enforced in the database. If a role needs wider reads, that's a governance amendment, never a config toggle.
Platform data defaults to a confidential class; each figure carries its class, source, and as-of so nothing is ever over-shared or mis-quoted.
Every figure change is recorded. The most sensitive classes never post to a channel, and answers above your clearance return a link, not the number.
Adoption is staged and safe by construction: the data layer first, then ingestion with a human in the loop, then query and dashboards, then the extras. Each stage earns the next.