Enterprise AI · Fund administration
Automate the investigation. Protect the ledger.
Xatva automates the work that consumes fund operations teams — heterogeneous custodian files, reconciliation breaks, exception investigation and NAV production support — while a deterministic financial engine controls every entry that reaches the ledger.
Onboarding early design partners for 2026.
app.xatva.ai — Exceptions · Meridian Global Fund
Reconciliation exceptions
NAV date 2026-06-30
Every stage writes to an append-only, hash-chained audit log.
$70tn+
in regulated open-end fund assets worldwide
100%
of AI investigations stored with full reasoning — evidence, hypotheses, conclusion
0
times AI writes directly to the ledger
The platform
Built for one of the world's most operationally complex industries.
Data Ingestion
Automatically understands heterogeneous files from custodians, brokers and administrators.
MT535 · MT536 · CSV · XLSX · PDF/OCR
Exception Investigation
AI investigates operational breaks while preserving complete reasoning and audit trails.
hypothesis → evidence → classification
Enterprise Governance
Deterministic financial controls, human oversight and enterprise-grade auditability.
sha-256 chained log · approval gates
Operating principle
Software that posts to a ledger must be deterministic. Software that investigates it can reason.
The costliest failure in fund administration isn't a slow close — it's a confident, wrong number reaching the NAV. Xatva is built so a model's mistake costs a review cycle, never a restatement.
Governance
Deterministic core. Probabilistic edge.
Language models never write to the ledger. Agents propose; a deterministic financial engine validates, posts and records. Every action — human or machine — lands in an append-only audit log.
append-only event log
Every event is SHA-256 hash-chained. Nothing is deleted or rewritten — corrections are new events, and history is permanent.
complete reasoning capture
Each AI investigation stores its full trace — evidence, hypotheses and conclusion — available to auditors and regulators on demand.
human approval gates
Material decisions require explicit sign-off. Confidence thresholds determine what reaches a human before anything is posted.
data boundary controls
Identifiers are tokenised before any model call. Sensitive client data never leaves the platform's control surface.
Timing
Why now?
Fund operations teams are asked to do more with the same headcount: custodian and broker data arrives in a dozen inconsistent formats, every reconciliation break demands manual investigation, and each manual touch is a chance for an error to reach the NAV.
Language models can now reason over that heterogeneous operational data — and deterministic systems can keep enforcing financial correctness while they do. For the first time, the investigation can be automated without loosening control of the ledger.
Models reason · Deterministic systems enforce
Request early access.
Xatva is onboarding a small number of design partners in fund administration. If that could be you, we'd like to talk.
Ronak Singh · ronrj01@gmail.com