Ascenzia Sentinel

The governance layerfor your AI agents.

Every call from your agents goes through the governance layer. You see, you decide, you prove.

See a demonstration on your case

Compliance screen: checks in place, AI system register, risk classification and exportable evidence logs

Spend tracking: spend for the period, projected total and a budget

Team AI usage: requests, people, sensitive exchanges detected and attempts blocked

List of AI modules in Ascenzia Sentinel, with their health: governed, to watch, to handle or stopped

Efficiency screen: share of requests served from cache, and the split by model class

The flow

Build. Govern. Route.

You build your agents in Ascenzia Studio. Ascenzia Sentinel governs every request. The gateway sends it to the right model.

Approved models

You choose.
They follow.

A catalogue of approved models. Each agent uses only the ones you have approved, according to the class of task.

Catalogue of approved models5 rules

Model	Class	Scope	Status
Large model	Reasoning	All teams	approved
Writing model	Writing	Legal, executive	approved
Fast model	Simple tasks	All teams	approved
Public model	Writing	DSI	limited
Unlisted model			blocked

Protect every request

Nothing gets through
without a rule.

Guardrails on the way in and out, an isolated run, connectors declared on a register.

01In and out

Per-request guardrails

Sensitive data spotted, requests outside the rules blocked.

02Run

Isolated run

Each agent runs inside its own boundary.

03Connectors

Connector register

An agent reaches only the tools you have declared.

Spend

Every budget, capped.

Spend tracked per agent, a maximum budget per agent, an alert before you reach it.

Budget per agentThis period1 alert

Spent65%

Cap100%

Alert threshold80%

Agent	Used	State
Council agent	69% of cap	ok
Quote follow-up	75% of cap	ok
Document sorting	83% of cap	alert
Client file	32% of cap	ok

Use of public AI30 days

Tools seen5

Sensitive data3.8%

Blocked attempts13

Tool	Relative use	Rule
Public chat		monitored
Browser assistant		monitored
Assistant		monitored
Work assistant		monitored
Unapproved tool		blocked

Shadow AI

See the AI
nobody declares.

Who uses public AI, and with which data. What must be blocked is blocked.

Model freedom

A model fails.
Your agents carry on.

A multi-model gateway and failover if a model fails. Changing model does not mean rewriting your agents.

Multi-model gatewayAutomatic failover

Agent	Model	Fallback	State
HR questions	Writing model	Large model	live
Bank matching	Large model	Writing model	live
Meeting summary	Fast model	Compact model	failover
Translation	Compact model	Fast model	live

Prove

Every action,
signed.

Signed traces and events: the material to show how you meet DORA, NIS2, the AI Act and GDPR.

Signed eventsChain verified

Time	Agent	Event	Signature
09:42:18	Council	Resolution read	a91f...3c
09:42:21	Council	Request to the large model	5d02...e7
09:42:22	Guardrail	Personal data masked	c7b4...11
09:43:05	Quote follow-up	Email draft created	0e6a...9f
09:44:40	Guardrail	External send blocked	f3d8...42

Already using a workplace assistant?

It stays.

Ascenzia adds governance.

Your tools stay in place. Ascenzia Sentinel adds the overview, the rules and the spend view across every agent and model.

Questions

Your questions about Ascenzia Sentinel

Can we cap the AI budget per agent?

Yes. Each agent has a maximum budget. Spend is tracked and an alert comes before the cap.

Can we change model without rewriting our agents?

Yes. The multi-model gateway separates your agents from the models. You change model, and failover is automatic.

How do we frame public AI assistants?

Shadow AI detection shows how your teams use public AI, and blocks what must be blocked.

See Ascenzia Sentinel

on your own case.

See a demonstration on your case hello@ascenzia.ai

Your AI agents. Steered. Secured. Sovereign.