Active pilot · Regulated financial institutions

Action control for agentic AI.

AIIAN routes AI-initiated payments, trades, procurement, and operational instructions through approval, risk, and audit controls before they reach real systems — giving enterprises the speed and confidence to act at scale with autonomous agents.

Agents propose. Enterprises authorise. Governance that raises the rate you can safely release — not a brake on autonomy.

Sits above your agent stack — Claude Code, LangChain, AutoGen, MCP, REST — no rip-and-replace. One control layer for every framework.

Request Invitation Try the demo

Built for consequential agent actions

Payments

High-value transfers and settlement instructions

Procurement

Purchase orders, suppliers, and spend workflows

Counterparties

Onboarding, screening, and risk-sensitive changes

Operations

API changes and business-critical system actions

One external control layer between agent frameworks and enterprise systems — with a single number for how close your agents are to an authorization run.

One control layer. Every agent framework.

Claude Code LangChain AutoGen MCP Server REST API Azure AI Foundry
DORA MiFID II Basel III J-SOX FCA

Built for your context

One control layer.
Three points of view.

Supervisory-ready by design

AIIAN does not assume agentic AI should be trusted because a model is trusted. It places an auditable, proof-bearing control layer between agent decisions and downstream mutation — built for supervisory review.

Proof-bearing, time-bounded release — every decision provable and replayable
FCA-aligned sandbox demonstrator — synthetic scenarios, no uncontrolled live-market exposure
Typed, machine-readable deficits, not opaque refusals
Request a walkthrough

Drop-in API

Make any agent action
releasable in a few lines.

Submit a proposed action; get back a signed release — or a typed deficit your agent can repair and resubmit. Then verify before you mutate.

The complexity — authority, evidence, capacity, reserve, audit — stays on our side of the API. You just call /execute and verify the proof.

Read the docs 
import { verifyBeforeMutate } from "@aiian/verifier-sdk";

// 1 · submit the agent's proposed action
let { baton, status } = await api.post("/v1/execute", {
  intent: "Settlement scenario — sandbox test",
  domain: "Finance", notional: 2_800_000, risk_level: 0.4,
});

// 2 · await the terminal decision (RELEASED | REJECTED | REVOKED)
while (status === "PENDING_REVIEW")
  ({ status } = await api.get(`/v1/execute/${baton}`));
if (status !== "RELEASED") return agent.repair(baton);

// 3 · verify the proof before you mutate state
const { proof } = await api.get(`/v1/execute/${baton}/evidence`);
if (verifyBeforeMutate(proof, { payload_hash }, { publicKeyPem }).ok) await settle();

The chain-of-command problem

AI agents don't just
recommend. They execute.

A Singapore clearing agent receives an instruction from a Tokyo orchestrator. It can verify the instruction came from another agent. It cannot verify that a human authorised it, that the amount is within limits, or that the counterparty passed the right checks — unless something upstream enforced those controls and left proof.

Four failure modes. All silent. All undetectable at the point of execution.

Identity spoofing — agent claims authority it was never granted
Instruction tampering — content modified between agents before execution
Replay — a prior approval reused against a new situation
Concurrent breach — two agents acting simultaneously exceed a shared limit

Tokyo · Orchestrator Agent

"Execute settlement · ¥420,000,000 · Counterparty XY"

Singapore · Execution Agent

Identity verified. But: sanctions? Limits? Human approval?

No way to know. No proof exists.

External Settlement System

Executed. Irreversible. £2,800,000 moved.

How it works

Enterprise approval
before AI execution.

Enterprise approval at the decision point — not inside the agent prompt, not reconstructed after the fact.

1

Connect your agents

Your AI agents connect to AIIAN via API or SDK. No changes to agent logic required. Works with any agent framework.

2

Apply enterprise controls

AIIAN applies your organisation's approval, risk, and compliance processes to each action before it reaches production. Policy stays with the enterprise.

3

Review outcomes

Approved actions proceed. Actions outside policy are stopped. Every outcome is available for internal review, audit, and regulatory response.

Policy is controlled by the enterprise — not embedded in the agent. Every decision is recorded.

What AIIAN provides

The governance layer
your agents need.

One layer above all agent frameworks. The same enterprise controls — regardless of what stack your agents run on.

Enterprise Control

Connect important AI-initiated actions to your organisation's control process before they affect real systems.

Risk Alignment

AIIAN helps ensure AI agent activity stays within the boundaries your organisation has defined — across all agents, not just some.

Audit Readiness

Every action — approved or stopped — is recorded for governance, regulatory, and internal review. Ready when you need it.

Multi-Agent Governance

Apply consistent enterprise controls across multiple AI agents, regardless of which framework or vendor they run on.

Human Oversight

Configure which actions require human review before proceeding. Reviewers see the full context and record their decision.

Sandbox Testing

Test realistic AI agent scenarios against your enterprise controls before connecting production systems. No production risk required.

The metric

Know your ρ.
The VaR for agentic action.

ρ (governability utilization) is the distance between the rate your agents act and the rate at which governance breaks — an authorization run. Covered means coverage ≥ 100%.

VaRdistance to a loss a desk cannot absorb
LCRdistance to a liquidity shortfall
ρdistance to an authorization run

ρ decomposes: it tells you whether risk rose from more agent activity, denser action chains, weaker observability, or thinner reserves — so you know which lever to pull.

AIIAN measures ρ, attributes it, and lowers it — by raising the rate you can safely release, not by slowing your agents down.

Governability utilization · example reading

ρ 0.41 COVERED

Coverage ratio (ACR) = 244% · runway ≈ 2.4× before the panic line.

This quarter's move

Agent rollout+20%
Stronger observability−5%
Added committed reserve−3%

Illustrative figures.

AIIAN Cloud Sandbox

Test before you
connect production.

Test realistic AI agent action scenarios against enterprise controls — without connecting production systems. Understand your control coverage before it matters.

  • Evaluate AI agent actions before they reach real systems
  • Run realistic scenarios against pre-configured enterprise controls
  • Review audit outcomes for every evaluated action
  • Integrate via API, SDK, or MCP — no agent logic changes required
Get sandbox access

Sandbox scenario · Settlement agent

Action

Large payment · Cross-border · New counterparty

Enterprise controls applied

Amount limit · Counterparty check · Approval routing

Approved

Audit recorded

Blocked

Audit recorded

Regulatory compliance

Built for regulated
environments.

DORA Art.28

ICT risk management for EU financial entities

MiFID II

Algorithmic trading controls & audit trails

Basel III

Operational risk & internal controls framework

J-SOX · FCA

Japan & UK financial services governance requirements

Designed to support audit, risk, and governance conversations in regulated environments.

Get started

If your agents act,
AIIAN connects them
to enterprise control.

Active pilot with regulated financial institutions. Settlement, FX, procurement, contract commitments — if your agents execute high-consequence actions, we want to talk.

Pilot access is by invitation. We respond to every request personally. No sales funnel.

Email hello@aiian.ai Try the live demo

Initial focus: financial institutions operating under DORA, Basel III, MiFID II, and J-SOX.