Private markets are projected to pass $30 trillion by 2030 (Preqin), and private credit alone to approach $5 trillion by 2029 (Morgan Stanley). Almost none of that data arrives in a clean feed. It arrives as PDFs: a partner capital account statement from one fund, a capital call notice from another, a distribution notice from a third, each in the general partner's own layout. Someone has to read them and key the numbers into a spreadsheet or an accounting system, every quarter, for every fund.
This guide is for the people doing that reading: LPs, family offices, fund-of-funds, allocators and the fund administrators who report on their behalf. It covers the documents, why the data is hard to get out, the tools built for the job, and where a self-serve tool like aiXtractor fits alongside the enterprise platforms.
The documents behind every private-markets portfolio
A single fund commitment generates a recurring stream of documents, and the figures in them feed valuation, reporting and tax:
- Partner capital account statement (PCAP). The definitive record of an investor's equity in a fund, usually issued quarterly. It shows the movement in capital over the period: contributions, distributions, income and loss, fees, carry, and the closing balance.
- Capital call notice. The GP's request to draw committed capital, with the amount, due date and the purpose of the call.
- Distribution notice. The return of capital or gains to investors, often split by type (return of capital, income, realised gain).
- Schedule K-1. The US tax form reporting each partner's share. Its capital figures are on a tax basis and can differ from the PCAP, which may be reported on a GAAP basis, so the two are not interchangeable.
- Fund financial statements. The longer, audited quarterly or annual accounts, including the schedule of investments and the notes.
Why this data is hard to get out
It is not an OCR problem. The difficulty is structural:
- Every GP formats differently. There is no universal layout for a capital account statement. A column one manager labels "Contributions" another calls "Capital called", and the roll-forward can run top-to-bottom or left-to-right. A tool tuned to one fund's format breaks on the next.
- It is PDF-first. The data is delivered as a document, not an API, and the volume grows every quarter.
- The quarterly cycle is brutal. Industry write-ups put the quarterly LP-reporting cycle at four to six weeks of senior staff time, and late data from the underlying funds is the most common cause of delay. The work is manual effort, not judgement.
- Spreadsheets have no audit trail. When figures are rekeyed across inboxes and spreadsheets, a single mistake can have a domino effect, and nobody can trace a reported number back to the statement it came from.
What is on a capital account statement
For the PCAP specifically, these are the fields that matter, and the ones a reconciliation has to hold together:
| Section | Fields |
|---|---|
| Identity | Fund, GP, investor name, account number, statement date, reporting period, currency |
| Commitment | Total commitment, contributed to date, unfunded commitment |
| Roll-forward | Beginning balance, contributions, distributions by type, allocated income and expense, management fees, carried interest, realised and unrealised gain or loss, ending balance |
| Position | Ending capital balance / NAV, period and inception-to-date columns |
The test of a correct extraction is simple: beginning balance, plus contributions, less distributions, plus income and gains, less fees and carry, should equal the ending balance. If it does not, a figure was misread.
The tools built for this, and where they leave gaps
There is a mature category of software aimed at this problem, mostly built for large institutional allocators:
- Canoe Intelligence is the best known. Its strength is collection: pulling documents automatically from thousands of GP portals and inboxes, then extracting the data and pushing it to platforms like Addepar. Built for LPs operating at scale.
- Accelex (now part of Carta) does extraction and analytics across private-market documents, with dashboards for IRR, MOIC and look-through metrics.
- Alkymi offers turnkey products for alternatives, including capital account statements and a dedicated private-credit tool.
- Cobalt AI Doc Ingest (FactSet), 73 Strings, Addepar Alts, Chronograph, Allvue and FundCount each tie extraction into their own valuation, accounting or reporting platform.
These are strong products. But they share a shape: enterprise platforms, sold through a sales cycle, with pricing that is not public, built around automated collection and a full data-management suite. If you are a large allocator who needs documents pulled from thousands of portals and loaded into a monitoring system, that is exactly right. If you just need the figures out of the statements in front of you, accurately and with something you can check, it is a lot of platform to buy.
Where aiXtractor fits
aiXtractor is the self-serve, API-first option for the extraction step itself:
- Every value is linked to its source. Click a figure in the result and the page highlights the exact cell it was read from on the PDF. For finance work, that is the difference between the extraction finishing and the extraction being right: a reviewer can confirm a reported number against the statement in seconds.
- The numbers are checked before you export. The capital roll-forward is reconciled (beginning, plus contributions, less distributions, plus income, less fees and carry, equals ending), unfunded commitment is checked against commitment less contributions, and dates are checked against the period. Disagreements are flagged, not silently exported.
- Any GP's format is a prompt, not a project. What to extract and what to check are written in plain language and editable per document type, so an unfamiliar layout is handled without waiting on a vendor.
- It is built to be integrated. A REST API with signed webhooks, and CSV, Excel or JSON out, so the figures drop straight into your model, accounting system or data warehouse.
Being honest about the boundary: aiXtractor extracts and verifies the documents you give it. It is not a collection platform that logs into GP portals for you, and it reads up to ten pages per file today, so a long audited fund financial statement is split into parts. For PCAPs, capital calls and distribution notices, the short, recurring statements that drive quarterly reporting, that is squarely the job it does.
Who it is for
Family offices and fund-of-funds reconciling statements from dozens of managers; smaller allocators and finance teams who cannot justify an enterprise seat; emerging managers and fund administrators who need clean, checkable data in a spreadsheet or an API response, this quarter, without a procurement cycle.
What users ask
Is my data secure? Documents are processed for extraction and are not used to train models. Results are yours to export or pull over the API.
Can it handle statements from different fund managers in one workflow? Yes. The extraction prompt is format-agnostic, so statements from different GPs run through the same pipeline and come back in the same structured shape.
What about a sixty-page audited financial statement? Split it into parts for now; the current limit is ten pages per file. The short quarterly statements, PCAPs, calls and distributions, are within that limit.
How do I know a figure is right? Two ways: the roll-forward reconciliation runs before export, and every value stays linked to its spot on the source PDF for a human to confirm.
How to try it on your statements
- Start on the free plan: 50 pages a month, no card.
- Upload a capital account statement, capital call or distribution notice as a PDF.
- Review the result: check the reconciliation flags and click any figure to see it highlighted on the page.
- Export to CSV, Excel or JSON, or wire the REST API into your reporting workflow.
The figures come back reconciled and traceable, so the quarterly rekeying becomes a review instead of a re-entry.