Cybervize
Agentic ISMS

Agentic ISMS: AI agents work inside your ISMS. You decide how far they go.

AI agents derive risks, measures and policies from your documents, fit new frameworks into what you already have and carry out tasks in the system. Every step remains traceable.

Book a free demo Book an intro call

60 minutes, live on demo data.

Definition

What is an Agentic ISMS?

An Agentic ISMS is an information security management system in which AI agents take part in the work: they derive risks, measures, controls and policies from an organisation's own documents, fit new requirements into what already exists and carry out tasks. A person sets how far an agent may act on its own, and every step remains traceable.

ISMS stands for information security management system, as described in ISO 27001.

Three modes

Three ways agents work in an ISMS

The AI in OdySecure works in three modes. The first two are available. Acting through connected agents is in trial with design partners.

  1. 1

    Derive

    Available in OdySecure

    You upload documents and provide details about your organisation. In six stages, this produces proposals for infrastructure, risks, measures, controls and policies. When a new framework comes in, the AI maps its requirements to what you already have and proposes what is missing. You accept or reject the proposals, one by one or per stage.

    See the ISMS module
  2. 2

    Ask and instruct

    Available in OdySecure, as the Navigator

    Ask your ISMS. Every answer comes with its source. On your instruction, the Navigator also does work in the system: it links a measure to a risk, creates tasks, records KPI values or schedules a BCM test. Nothing runs until you have confirmed its plan, and it only acts with your permissions.

    See the Navigator
  3. 3

    Act

    Design-partner programme, not yet generally available

    Connected agents work under their own identity, with their own role and an autonomy level that you set per task, for example in incident triage. An agent can never close an incident, accept a risk or grant itself permissions.

Ask and instruct

What the Navigator does on your instruction

You write in the chat what needs doing. The Navigator puts forward a plan and carries it out only after you confirm, with your permissions. You can usually undo a confirmed action within the next two hours.

  1. 1

    Instruction

    For example: “Create a task to review the password policy.”

  2. 2

    Plan

    The Navigator shows which entries it will create or change.

  3. 3

    Confirmation

    Only now is it carried out. Anything your role may not do is never offered.

More about the Navigator

Examples from the twelve actions

  • Link a measure to a risk, with the residual risk recalculated
  • Assign an asset to a business process
  • Change the treatment, rationale or owner of a risk
  • Create a task for a policy, a risk or a supplier
  • Record a KPI value for a period
  • Schedule a BCM test for a contingency plan
  • Start a due diligence review for a supplier
  • Request a risk acceptance or a time-limited exception. People decide on it.

Which actions it offers depends on your role and the modules you have licensed.

Limits

You decide how far an agent may go

Duties and responsibility for information security remain with your organisation and its management, even when agents take part in the work. For connected agents, you set an autonomy level per task. As soon as an agent carries out work itself, the task needs a person who is accountable for it.

Autonomy levels for connected agents

  1. A0

    Observe

    The agent reads and reports.

  2. A1

    Suggest

    It makes suggestions, and a person takes them over.

  3. A2

    Prepare

    It prepares the work, and a person completes it.

  4. A3

    Reversible action

    It carries out changes that can be reversed.

  5. A4

    Action after approval

    It carries out work once a person has approved it.

  6. A5

    Prohibited for agents

    These actions are reserved for people.

  7. A6

    Autonomous up to the risk threshold

    It acts on its own as long as consequences and reach stay below your threshold. Above it, it needs approval.

Connected agents are in trial in the design-partner programme. Levels at which an agent carries out work itself can only be approved by a second person.

What always stays with people

  • Proposals from the derivation take effect only once a person accepts them.
  • Whatever the Navigator changes, it carries out only after your confirmation.
  • Whether a risk is accepted is decided by your organisation, never by an agent.
  • A person closes an incident.
  • Your administrators grant permissions. An agent cannot grant itself any.
  • In a vCISO mandate, a senior CISO from Cybervize leads the implementation and keeps your management informed.

Comparison

Classic ISMS software and an Agentic ISMS

Both run the same management system. What differs is who does the work. The middle column describes the usual shape of ISMS software, not a specific product.

QuestionClassic ISMS softwareAgentic ISMS
Where do risks and measures come from?From questionnaires and catalogues that someone in-house works through.The AI derives them from your documents. You accept the proposals one by one or per stage.
Where do the policies come from?From templates you adapt to your organisation.The AI drafts them from what your documents say. Every draft goes through review and approval.
What happens with a new framework?Another catalogue that someone works through alongside the existing one.The AI maps its requirements to your existing risks and measures. Where something is missing, it proposes it.
How do you answer “Where do we stand?”With a report or an exported spreadsheet.Ask your ISMS. Every answer comes with its source.
Who does the routine work in the system?Your team, entry by entry.Agents, on instruction or as a fixed task in a security process, within the limits you set.
Who approves?Approval workflows for documents; anything else applies once it is saved.You decide what an agent may do on its own and what needs approval. Proposals remain proposals until someone accepts them.
How is the evidence created?Often as a separate task shortly before the audit.In day-to-day operation: changes in the audit log, with time, editor and old and new value.
What does your security officer work on?Creating and maintaining entries.Reviewing and deciding.

What an ISMS has to achieve is the same in both columns.

Evidence

How you show what happened in the ISMS

For an audit, what counts is who approved what and when. That applies equally to entries made by people and to the work of the AI.

  • OdySecure logs changes with the time, the editor and the old and new value.

  • For AI proposals, OdySecure stores the model, the reasoning, their supporting passages in the documents and who reviewed them.

  • OdySecure logs questions to the Navigator in your organisation's audit log, and its actions together with the person who confirmed them.

  • The log can only be appended to and is sealed every hour as a hash chain. A verification package lets you recompute the chain outside OdySecure.

  • Clause 7.5.2 of ISO 27001 requires documented information to be reviewed and approved. In OdySecure, policies go through drafting, review and approval, by default with two different people.

The model behind it

Why agents can work in OdySecure

An agent can only derive what is connected. In OdySecure, business process, asset, risk, measure, control, policy and security process sit in one data model. We call it the ISMS operating system. Through these links, a proposal can be traced back to the requirement it came from.

More about OdySecure
One linked data foundation: from the business process to the security processBusinessprocessAssetRiskMeasureControlPolicySecurityprocess

Getting started

How to get an Agentic ISMS

OdySecure comes in two ways. The AI works the same in both.

  • In a vCISO mandate

    A senior CISO from Cybervize leads the implementation of NIS-2, DORA or ISO 27001 in OdySecure. The mandate starts at €3,600 a month, with the licence included.

    See the vCISO mandate

  • As a licence

    Your team runs OdySecure itself, with an implementation project if you wish. Prices per tier are on our pricing page.

    See pricing

Frequently asked questions

What is an Agentic ISMS?
An Agentic ISMS is an information security management system in which AI agents take part in the work: they derive risks, measures, controls and policies from an organisation's own documents, fit new requirements into what already exists and carry out tasks. A person sets how far an agent may act on its own, and every step remains traceable.
How does an Agentic ISMS differ from Agentic GRC?
Agentic GRC usually refers to tools in which AI agents handle individual compliance tasks, such as collecting evidence or filling in questionnaires. In an Agentic ISMS, the agents work inside the management system itself: they build the model of risks, measures and controls, fit every new framework into it and take on tasks in the security processes, up to incident triage. Evidence for ISO 27001, NIS-2 or DORA comes out of day-to-day operation.
Can the Navigator also change data?
Yes, on your instruction. For example, it creates tasks, links measures to risks or records KPI values. It shows you its plan first, and nothing runs until you confirm, with your permissions. It cannot accept a risk; it can only file the request.
How does an Agentic ISMS take on a new framework?
The AI maps the requirements of the new framework to the risks, measures and controls that already exist in your ISMS. Where none fits, it proposes a risk and a measure, and you decide. You keep maintaining one model, even with several frameworks such as BSI IT-Grundschutz, NIS-2 or DORA.
Does an Agentic ISMS replace the CISO or the information security officer?
No. Agents do work within the limits your organisation sets. Approvals, accepting risks and closing incidents stay with people, and responsibility stays with your management. In a vCISO mandate, a senior CISO from Cybervize leads the implementation.
Is an Agentic ISMS compatible with ISO 27001?
Yes. ISO 27001 sets out what an ISMS has to achieve and does not prescribe a tool. Clause 7.5.2 requires documented information to be reviewed and approved. In an Agentic ISMS, this approval is a fixed part of the process. In OdySecure, policies go through drafting, review and approval.
What happens if an agent gets it wrong?
Then you reject the proposal. Proposals from the derivation apply only once you accept them. If the Navigator changed something on your instruction, you can usually undo it within the next two hours. You can cancel a derivation run and resume it later.
Where do the AI models run?
For OdySecure, by default on Cybervize's own infrastructure, without external model providers. If you wish, you can connect your own model for your tenant.
When will connected agents in OdySecure be able to act on their own?
The first two modes, Derive and Ask and instruct, are available today. Connected agents with their own identity, for example for incident triage, are being trialled in the design-partner programme. They are not yet generally available. If you want to use them early, you can apply as a design partner.

See an Agentic ISMS in the demo

In 60 minutes we show you OdySecure on demo data: how a derivation run works, how the Navigator answers and carries out instructions, and where you approve. If you would rather talk about a vCISO mandate first, book an intro call.

Book a free demo Book an intro call