Aevus is the neutral, cross-system layer that lets private funds run their back office with agents. A human only steps in to approve.
Capital calls, NAV close, LP reporting, and investor onboarding are still assembled by hand. Analysts pull data from fund admin, custodians, data rooms, and accounting systems, then re-key it into spreadsheets. Agents could do this work, but a language model cannot safely move a wire, run a waterfall, or update an LP's capital account on its own.
An agent asks for an outcome, like issue a capital call, not to log into Allvue and click twelve things. Aevus does the rest, and the firm sees one clean, governed action instead of a dozen systems. Hiding fragmentation is the product.
An agent asks for an outcome, like issuing a capital call. It never logs into a portal or touches a screen.
Aevus does the math against the firm's source of truth. The agent never freehands a number, a wire, or an allocation.
Every action is checked against the firm's own rules, then held for a human to approve before anything happens.
The approved action is carried out a single time and written to a permanent audit record: what happened, and who approved it.
Agents act on what the firm wants done, not where it lives. Hiding fragmentation is the product.
The agent plans and requests; the engine computes and executes. No number, wire, or allocation is ever freehanded by a model.
Limits, thresholds and segregation of duties are enforced when a tool is called, not in a prompt. This is what lets a firm trust any agent.
Anything that moves money, files externally, or touches an LP passes a configurable approval gate.
Every action runs a single time and writes a permanent, exportable record with full provenance. Machine speed, ledger-kept.
Aevus works across the systems a firm already runs, whoever built them.
Two workflows where Aevus earns its place fast: LP reporting and capital-call prep. Both run end-to-end across every system they touch, and both deliver value without moving a dollar, so your team can trust the work before anything irreversible happens.
Produce and distribute periodic investor reporting: capital-account statements, quarterly reports, ILPA-aligned templates and ad-hoc LP packages, assembled per LP and honoring every side letter.
Prepare the whole capital call right up to the wire: per-LP allocations pro-rata to unfunded commitment, honoring excuse, exclude and side letters, with ILPA notices drafted and the package ready for sign-off. Computed deterministically; your team sends the cash.
A firm gates entirely on whether it trusts Aevus near its systems. So security is designed into every action from the start, not bolted on. These are the principles we build to.
Aevus computes and validates every money action. The model never freehands a number, a wire, or an allocation.
Limits, dual control, segregation of duties and approval gates are enforced by the engine when a tool is called, never by a prompt.
Every action writes an immutable, exportable record: source, computation, approver and timestamp. Built for LPs, auditors and regulators.
Credential vaulting, hard tenant isolation, least-privilege agent identity and SOC 2-aligned controls. We will walk your security team through the architecture.
We work closely with a small number of firms putting agents safely to work in the back office. If that is where you are headed, let's talk.