NeoXam Agents — launching Q3 2026
AI where your investment work already lives.
Governed, auditable AI agents embedded in the platforms that already run your investment operations.
Built by NeoXam — 150+ financial institutions, €25tn+ assets on our systems.
- financial institutions
- 150+
- financial institutions
- assets on NeoXam core systems
- €25tn+
- assets on NeoXam core systems
- running critical investment workflows
- 35 years
- running critical investment workflows
- worldwide, headquartered in Paris
- 21 offices
- worldwide, headquartered in Paris
Why now
From copilots to governed agents
The copilot wave is over — every vendor has one. Answers are not outcomes, and regulated institutions cannot run what they cannot certify.
01
Copilots answer questions
Copilots sit beside your tools and reply in prose. They do not investigate a break, validate a business rule, or pre-work a NAV exception.
02
Pilots stall at the production gate
78% of investment managers are piloting agentic AI; only 27% report significant impact (SimCorp, 2025). The missing piece is not a better model — it is governance.
03
Agents, governed by construction
Every agent is declared and versioned, gated by evaluation baselines, policy-controlled at every tool call, and traced end to end. Humans stay in command.
Why NeoXam Agents
Built for production, not for demos
Four commitments, each backed by architecture — not by a roadmap slide.
Embedded where the work happens
Copilots sit beside your tools. NeoXam Agents work inside them.
Agents are invoked from the screens your teams already use and return typed, schema-validated outputs your workflows consume directly. No new destination, no workflow migration.
Governed and auditable by design
Compliance-grade by construction, not retrofit.
Reconstruct every run: model, ordered tool calls, tokens, cost, duration — plus the exact catalog version that executed. Quality is gated by evaluation baselines that can only go up.
Your deployment, your rules
EU by default. Sovereign by choice. Never married to one model.
EU (Frankfurt) deployment by default and open standards throughout — MCP, OIDC, OpenTelemetry. The model is a field in each agent's descriptor: switching providers is a configuration change, not a re-architecture.
Built on buy-side DNA
The agent layer above systems already running €25 trillion.
Agents speak ABOR/IBOR, breaks, business rules and NAV — querying the same governed data models your operations already trust, through product-owned MCP servers.
Platform
Capabilities, honestly labeled
We publish the status of every capability and never present roadmap as product.
Agent catalog & governance
ShippedEvery agent, tool and skill declared in a Git-governed catalog. All changes go through PR review — nothing undeclared can run.
Task-mode runtime
ShippedOne production-grade API for every product: typed job in, schema-validated output back, live progress, cancellation, run history.
Structured outputs
ShippedTyped, schema-validated results your workflows consume directly. A malformed answer is a hard failure, never a confident guess.
Observability & audit trail
ShippedAn OpenTelemetry trace for every model and tool call, and an append-only audit log designed for 7-year regulatory retention.
Cost governance
ShippedHard per-run budgets — duration, tool calls, tokens — with per-tenant quotas and cost dashboards per tenant, agent and run.
Eval gates
ShippedVersioned evaluation datasets and baseline scores that can only rise. Advisory in Beta, blocking at GA.
Human-in-the-loop approvals
BetaPer-tool allow/ask/deny policies. Sensitive actions pause for explicit, recorded human approval.
Conversational agents
BetaMulti-turn assistants embedded in product screens, with platform-managed memory and streamed answers.
Products
Agents across the NeoXam suite
Embedded in the systems of record your teams already run — with statuses published, never blurred.
- Shipped
Aro
Reconciliation
Every break investigated before your analyst opens it — root cause, evidence, recommended action, fully traced.
Pilot delivered end-to-end
- Shipped
DataHub
Data management
Business rules drafted, validated, evaluated and tested by agents that know your syntax dictionary and rule corpus.
Task mode; assistant in Beta
- Coming
GP
Fund accounting
NAV exceptions pre-investigated with audit-grade evidence, so your close gets faster without losing oversight.
GA wave
- Coming
Impress
Reporting
From data to a drafted, contextualized report — query generation, template drafting and narrative context.
GA wave
- Coming
Density
Portfolio & compliance
Compliance breaches explained with traceable evidence at the moment they are flagged.
GA wave
Trust
Governance your auditors can read
Auditability is the design center, not a feature flag. Every claim below maps to an architectural mechanism.
Append-only audit trail
ShippedEvery run reconstructable step by step — designed for 7-year regulatory retention and exportable for compliance.
Evaluation gates
ShippedEach agent ships a versioned eval dataset and a baseline score that can only go up. Advisory today, blocking at GA.
Human-in-the-loop
BetaSensitive actions pause for explicit human approval — who, when and why recorded as first-class audit entries.
EU data residency
ShippedEU (Frankfurt) deployment by default. Client business data stays inside your tenant boundary; region pinning lands at GA.
OIDC SSO
BetaEnterprise single sign-on — KeyCloak, Entra ID, Okta-compatible — with passwordless magic-link sign-in today.
Tenant isolation & RBAC
ShippedRow-level isolation enforced in the database, per-agent roles and scoped credential vaults — secrets never reach prompts, logs or traces.
Move from AI pilots to audited production
The Early Adopter Program is open now: a limited cohort, a one-year platform trial, and three workshop streams — Business ROI, Compliance, and Operational fit. General availability comes in Q3 2026.
Platform console — conversational agents arrive with the V1 Beta.