AKRU's document intelligence reads asset agreements, security and credit contracts, and loan documents and converts them into structured, actionable data — so the terms that govern a fund flow straight into the platform instead of being re-keyed by hand. Natural-language processing and large language models, applied to the documents institutions actually run on.
LPA extraction of waterfall provisions, fee terms, share classes, and transfer restrictions — the economic engine of the fund, read from the source.
Subscription documents
Investor details, commitments, eligibility representations, and elections pulled from sub docs into onboarding and the cap table.
K-1s
K-1 document automation in both directions: incoming K-1s parsed into structured allocations, and outgoing K-1s generated from the same platform record.
Loan agreements
Facility terms, covenants, rates, and payment schedules from loan and credit documents — structured for servicing and monitoring.
Financial statements
Statement line items extracted into comparable, structured data for oversight and reporting across positions.
Extraction to action
Extraction feeds the event fabric — with a human at the gate
Extraction alone just makes a smarter filing cabinet. What makes document intelligence operational is where the data lands: extracted terms populate the platform's event fabric directly, so the LPA's waterfall provisions become the waterfall engine's configuration, the sub doc's commitment becomes the LP's cap-table position, and the loan agreement's payment schedule becomes servicing events. When a term changes upstream, the fabric carries it downstream — nothing is re-keyed twice.
And nothing is trusted blindly. Every extraction passes through human-in-the-loop QA gates: fields below a confidence threshold are flagged for reviewer confirmation, and economically significant terms — waterfall tiers, fee provisions, transfer restrictions — are verified by a person before they drive a calculation. Automation does the reading; accountability stays with a reviewer. That's the same discipline as the platform's $0.01 K-1 QA gate, applied at the point of ingestion.
Questions teams ask
Frequently asked questions
What document types does AKRU's document intelligence support?
Limited partnership agreements (LPAs), subscription documents, K-1s, loan agreements and credit documents, security contracts, and financial statements — the paperwork institutional funds actually run on.
Is the extraction fully automated, or does a human review it?
Both, deliberately. Extraction is automated; acceptance is gated. Extracted terms pass through human-in-the-loop QA gates — low-confidence fields are flagged for review, and economically significant terms like waterfall provisions are verified before they drive downstream calculations.
Where does the extracted data go?
Into the platform's event fabric. Extracted terms populate the cap table, waterfall engine, and tax suite directly — so the LPA's actual terms drive allocations, distributions, and K-1s without manual re-keying.
Put your documents to work
A 15-minute look at document intelligence against your own paperwork.
