Active deployment · 2026

Fund administration, rebuilt from primary sources.

NAV packages today are humans gluing together workflow tools. Aktus deploys bespoke workers that read your LPA, parse your trial balance, and assemble waterfalls, every figure traced to the document it came from.

14 days
From scoped pilot to first live workpaper
92%
Of NAV components produced without human touch
SOC 2 Type II
Sovereign deployment in your VPC
6x
More funds per controller
/ 01
Who we serve

Built for the funds running the alts stack today.

Aktus deploys inside the firms whose NAV packages, capital calls, and investor reports are the operational fabric of private capital.

Fund Administrators

Outsourced fund admins

Top-25 admins running thousands of funds across vehicles, currencies, and waterfalls.

  • NAV production at scale
  • Cross-fund consolidation
  • Investor reporting automation
In-house Teams

GP-side admin teams

Sponsors who took admin in-house for control, now staring at headcount math.

  • LPA-driven workpapers
  • Capital activity packages
  • Waterfall calculation review
Hybrid

Co-sourced operations

Mid-market firms where the GP keeps the close and the admin keeps the books.

  • Trial balance reconciliation
  • Allocation engine outputs
  • Tie-out to source documents
/ 02
What Aktus automates

The work that consumes the close.

A bespoke AI worker isn’t a chatbot, it’s a system that reads the same documents your team reads and produces the same artifacts, on a deterministic schedule.

Inputs our AI worker ingests
  • Limited Partnership Agreements & side letters
  • Trial balances and general ledger extracts
  • Capital activity statements from custodians
  • Subscription and transfer documents
  • Manager statements and broker confirms
  • Historical workpapers for prior-period tie-out
Outputs our AI worker produces
  • NAV workpapers with line-level source citations
  • Capital call & distribution notices, per-LP
  • Carried interest and waterfall computations
  • Management fee accruals against LPA terms
  • Investor capital account statements
  • Audit-ready binders with full provenance
What stays with your team: sign-off, judgment calls, and the relationships. Our AI worker doesn’t replace the controller, it removes 70% of manual document handling.
/ 03
How it works

Five steps from your LPA to a live workpaper.

Deployment is a sequence, not a platform install. Each step is gated on your sign-off.

01

Scope

We pick one fund and one artifact, usually NAV, and define the source-of-truth documents.

Scope exampleA $400M direct lending fund · Q3 NAV package.
02

Ingest

Your documents land in a sovereign environment. Our AI worker indexes them with line-level provenance.

ExampleLPA, side letters, GL, custodian statements, prior workpapers.
03

Calibrate

We tune the AI worker against three prior periods you’ve already closed. Variance reports surface every disagreement.

ExampleWorker reproduces Q1 NAV within $0.04 / LP unit.
04

Deploy

Our AI worker runs on your close cadence. Outputs land in your audit binder with one-click trace to source.

ExampleLive in fund #1. Roll-out plan for next 5.
05

Expand

New funds onboarded by configuration, not engineering. Our AI worker generalizes across LPA variations.

ExampleFund #6 onboarded in 4 days.
You sign off on the AI worker before it produces a single LP-facing artifact. No deployment is a black box.
/ 04
Why this needs Aktus

Frontier models can’t do fund admin alone.

Aktus makes them production-ready for Alternative Assets

The frontier model problemFund administration is not a reading comprehension problem, it’s a long-running, side-effect-laden production process with non-negotiable provenance requirements. A model that hallucinates a single basis point is unusable; a model that can’t explain where a number came from is unauditable.
What Aktus solves

Source-anchored reasoning

Every output cites the document, page, and clause it derives from. The worker refuses to produce a figure it can’t trace.

Verified by an evaluation harness on every run.
Where models alone fail

Multi-document long-context

A NAV package crosses 40+ documents, hundreds of pages. Frontier context windows blur the distinction between them.

Aktus uses a retrieval-first architecture.
Worse than humans

Side-effect operations

Booking journal entries, sending capital calls, posting to a GL, none of these tolerate the failure modes of a chat interface.

Aktus runs deterministic, audited tool calls.
An AI worker is not a model. It is a model wrapped in a process, governed by an evaluation harness, and run inside your perimeter, built bespoke for one job.
/ 05
Context orchestration

How our AI worker stays accurate.

Aktus runs every worker against a continuously evaluated context graph, your LPAs, your prior closes, your house policies, so the worker’s answers track your firm, not the public internet.

01

Document graph

Every primary source, LPA, side letter, GL, is parsed into a typed graph with version history and effective dates.

02

Policy layer

Your house interpretations of LPA terms (waterfall tiers, expense definitions) are first-class objects the worker queries.

03

Eval harness

A regression suite of prior-period closes runs on every model and prompt change. Deployment is gated on green.

04

Human review

High-impact outputs route to a named human reviewer. Their edits update the policy layer for next time.

/ 06
Economics

A unit cost change, not a tooling change.

Fund admin is priced per fund, per investor, per period. Aktus changes the cost structure underneath that price, not the price itself.

Admin economics today
  • ×Headcount scales linearly with funds onboarded
  • ×Senior reviewer time consumed by line-level tie-out
  • ×Quarterly close is a peak-load staffing problem
  • ×New fund types require new playbooks and training
  • ×Margin compression every renewal cycle
With Aktus inside
  • Headcount scales with relationships, not artifacts
  • Reviewer time is exception-driven, not line-by-line
  • Close cadence flattens, same effort across the quarter
  • New fund types are a configuration, not a hire
  • Per-fund unit cost falls each quarter the worker runs
70%
Reduction in time per NAV close on funds in production
More funds per controller, without sacrificing review depth
< 90 days
Payback on a typical 4-fund deployment
/ 07
Sovereign deployment

Your perimeter. Your data. Your governance.

Aktus runs inside your environment, against your documents, under your audit trail. No data leaves the boundary you control.

Sovereign by default

Deployed in your VPC or on-prem. Aktus operators see logs and metrics, never primary documents.

  • SOC 2 Type II
  • AWS, Azure, GCP, on-prem
  • Customer-managed keys

Audit-ready by design

Every worker output carries a hash-chain back to the document, page, and clause it derives from.

  • Immutable provenance log
  • Reviewer sign-off captured
  • Exportable for Big-4 audit

Model-agnostic

Run on frontier APIs, your private inference, or hybrid. Switch underlying models without re-deploying the worker.

  • Anthropic, OpenAI, on-prem
  • Per-task model routing
  • Versioned, evaluated upgrades
Compatible with your alts stack
Allvue
Investran
eFront
Workiva
Carta
Sungard

Bring Aktus inside your fund admin team.

Your fund admin expertise. Our AI engine. Software-like margins on a services business.