Security & control

Control isn’t added after AI.
It defines what AI is.

MyCh IQ makes operational boundaries visible: what can be executed, what needs approval and what remains human.

Discuss your framework ↗

Four layers of control

A clear boundary at every step.

The goal is not to prevent automation. It is to make its scope explicit and verifiable.

Scope

Approved tools

The AI agent acts only through defined actions, with expected inputs and an explicit scope.

Autonomy

Graduated autonomy

Each action can be supervised, subject to approval or executed autonomously according to its policy.

Accountability

Roles and permissions

Responsibilities, thresholds and the scope of approvals are assigned to identifiable roles.

Traceability

Decision audit

Conversation, decision, actor, tool, approval and outcome form a usable audit trail.

NÉNovera Énergie⌄

Demonstration dataCL

AI SUPERVISION / AUTONOMY

Autonomy and permissions

Supervised

Every action is approved

Semi-automaticActive

4 autonomous actions

Autonomous

Subject to each action’s policy

Action	Policy	Limit	Scope
Retrieve an invoiceBilling`recuperer_facture`	Autonomous	—	Read only
Check a meter readingBilling`consulter_releve_compteur`	Autonomous	—	Read only
Send confirmationCommunication`envoyer_confirmation`	Autonomous	—	Customer’s language
Close a ticketProcess`cloturer_ticket`	Autonomous	—	After resolution
Change customer addressCustomer record`mettre_a_jour_adresse_client`	Approval	—	Contract address
Schedule a visitField service`planifier_intervention`	Approval	< 30 d	Service area
Offer goodwill creditFinancial`geste_commercial`	Approval	≤ €100	Supervisor
Issue a refundFinancial`rembourser_client`	Human only	Any amount	Human only

Each action’s policy takes precedence over the global mode.

A human-only action never becomes autonomous.

Risk exists at the action level

A different policy for every decision.

Retrieving an invoice, changing customer data and triggering a financial transaction carry different risks. MyCh IQ lets you govern each separately.

Autonomy for simple, sufficiently documented actions
Approval for sensitive changes
Mandatory human control when policy requires it

Traceability

Understand what actually happened, afterwards.

The audit shows decisions and their operational context without exposing secrets, tokens or infrastructure details in the business interface.

NÉNovera Énergie⌄

Demonstration dataCL

AI SUPERVISION / AUDIT LOG

Audit log

20–21 August 2026

All actors8 entries of 1,284

Timestamp	Decision	Actor / tool	Result
21/08 11:24:38`CONV-4821`	Autonomous resolutionAI agent	AI agent`consulter_releve_compteur`	Invoice corrected`TR-8f41c2`
21/08 11:22:54`CONV-4818`	Contextual handoffAI agent → Y. El Amrani	AI agent → Y. El Amrani	Transferred with context`TR-8f4197`
21/08 11:20:11`CONV-4815`	Approval requestedAI agent → C. Laurent	AI agent → C. Laurent`geste_commercial`	Pending`TR-8f4163`
21/08 09:41:04`CONV-4802`	Action approvedCamille Laurent	Camille Laurent`mettre_a_jour_adresse_client`	Record updated`TR-8f40a7`
21/08 09:12:20`CONV-4796`	Approval declinedCamille Laurent	Camille Laurent`geste_commercial`	Response adjusted`TR-8f4055`
20/08 17:48:02`CONV-4771`	Autonomous resolutionAI agent	AI agent`recuperer_facture`	Invoice sent`TR-8f3ee1`
20/08 16:31:47`CONV-4762`	Action failedAI agent	AI agent`planifier_intervention`	Retry scheduled`TR-8f3e70`
20/08 15:02:19`CONV-4750`	Policy changedCamille Laurent	Camille Laurent	Address → approval required`TR-8f3dc4`

24-month retention · EU regionNo tokens or secrets logged

A deliberately specific scope

This page describes the controls visible in the MyCh IQ demo. Hosting, retention, integration, GDPR and governance requirements are validated with each organisation during scoping. No unconfirmed certifications are claimed here.

Start with the rules

Show us what AI must never decide alone.

We’ll build the demo around that boundary, then around the actions you actually want to automate.

Book a demo ↗