Find a process
We sit with the people doing the work and map where the hours go: portal logins, spreadsheet cleanup, copied transactions, checks, and follow-up emails.
AI AGENTS FOR FUND OPERATIONS
AI agents, delivered and managed.
Instead of giving your team another software login, Peregrine delivers the completed operational task. We map the process, build the agent, connect it to your existing stack, and maintain it while it runs.
Built around your processControlled accessManaged after launch
HOW IT WORKS
Three steps from manual work to a task that runs itself.
Build and deploy the agent
The agent is built around your formats, controls, approval points, and systems. There is no forced platform migration or generic workflow to adopt.
Let it run
The process runs on schedule, records every action, and sends clean output or exceptions to the right person. Peregrine monitors and improves it over time.
THIS IS NOT A CHATBOT
Your team reviews the output instead of producing it.
Chat tools can help someone write or analyze. An operational agent can be granted narrowly scoped access, a measurable task, and permission to complete approved steps end to end.
Chat assistantPrompt driven
Conversation
“Upload the statement and tell me what you need.”
The next action waits for a person.Useful for assistance, but the operational responsibility remains with your team.
Managed operational agentOutcome driven
ScheduledNo prompt required
- ✓Email and file sources checked
- ✓Data transformed to the target format
- ✓Controls applied and exceptions flagged
- ✓Result delivered to the approved system
The task finishes inside agreed boundaries and the run remains traceable.
WHERE WE OFTEN START
The NAV lifecycle.
These are four representative agents across a NAV workflow. The same design pattern applies anywhere data moves manually between systems.
Ingest agent
Pulls files and data from brokers, banks, inboxes, APIs, and supported onchain sources on schedule.
Manual logins and exports are reduced.Structure agent
Reads inconsistent statements and converts them into structured records in the exact layout the destination platform expects.
Daily transcription becomes a controlled transform.Push NAV agent
Loads validated records into the accounting or NAV system, collects results, and prepares the review package.
Copy-and-paste and upload errors are reduced.Tokenize agent
Feeds approved NAV events, distributions, and unit-register changes into the tokenization layer through controlled calls.
Offchain operations can connect to onchain records.MORE AGENTS
Operational agents beyond NAV.
The right first agent depends on volume, repeatability, data availability, and how costly exceptions are today.
Compliance agent
Monitors defined regulatory sources and internal obligations, then routes relevant changes or overdue evidence to the right owner.
Document agent
Reads legal documents, term sheets, subscription packs, and reports to extract structured terms and highlight inconsistencies.
Risk agent
Checks concentration, exposures, stale prices, limit breaches, and unusual patterns against an agreed rule set.
Communication agent
Drafts investor updates, recurring reports, and operational responses from approved data and style guidelines for human review.
HOW WE BUILD
Built around your process and managed after launch.
Wired into your stack
The agent works with approved custodians, admin platforms, banks, trading venues, inboxes, drives, and data providers.
Bespoke, not off the shelf
Inputs, schemas, controls, approvals, logging, and outputs are designed around the actual task rather than a generic automation catalog.
Guardrails that hold
Access is scoped, credentials are separated, actions are logged, thresholds are explicit, and unusual cases stop for review.
Managed by Peregrine
We monitor the implementation, handle failures, update connectors, and improve the process as systems and requirements evolve.
PRIVATE & LOCAL AI
Need it fully private? Run the agents locally.
Some organisations cannot send investor data, positions, or documents to a public model provider. A local deployment can use open-source models inside your infrastructure, with the same workflow controls and audit trail.
Discuss a private deployment →On your infrastructure
Models and agents can run on your servers or private cloud under your access, monitoring, and network policies.
No public AI API required
Sensitive data can remain within the defined perimeter rather than being transmitted to a third-party model endpoint.
Same operational controls
Approval gates, run logs, exception routing, and human oversight remain part of the implementation regardless of where the model runs.
THE PEREGRINE STACK
Every layer can be operated programmatically.
Agents can use the planning, tokenization, settlement, and registry layers as controlled building blocks in a broader fund workflow.
01
Fund Builder
Map legal, administration, custody, audit, and management-company requirements.
02Tokenization Engine
Issue and administer programmable fund units with a live register.
03OmniClear
Connect fiat and stablecoin rails for subscriptions and redemptions.
04Fund Registry
Publish searchable fund profiles for investors, allocators, and machine-readable discovery.
WHICH PROCESS SHOULD GO FIRST?
Show us the manual work. We will identify the first useful agent.
The best starting point is usually high-volume, repeatable, deadline-driven work with clear inputs, outputs, and review rules.