Why you approve every change

Nothing your AI drafts reaches your compliance record without an approval from a signed-in person's account. Whatever it drafts is checked on our servers first. It then waits until someone signed in to your workspace approves it, with a click on the single-use approval link or in the dashboard, and the approval is recorded against that person's account. Asking in the chat is never enough. Kanonik cannot tell whether the person, or software acting in their signed-in browser, made the click. A few things are saved without an approval; they are listed below.

The approval gate is not a setting you can turn off, and it is not something your AI can work around. This article explains how the gate works and what your part in it is.

This works the same way for every framework Kanonik supports. It is about how a change reaches your record, not about any one standard.

Why a human approves every change

Your compliance record is only as good as its defensibility. When an auditor or a customer reviews it, the question behind every entry is the same: who stood behind this, and can you prove it. The approval gate answers that question for every change. A named person's signed-in account approved the change, and that act is recorded against them.

That is what keeps a human accountable for each entry. The AI does the drafting and the legwork, but it does not get to write to your record on its own. You do, by approving. This is what makes the record hold up later, rather than being a pile of changes no one signed for.

How approval works

A change moves through a fixed sequence, and you enter at the end:

1. Your AI prepares the work. It drafts the policy, the control, the mapping, or whatever you asked for, and submits it.

2. The safety check runs first. A check runs on Kanonik's servers and reaches a verdict on each item before anything is offered to you. You are never asked to decide on something the check has not looked at.

3. You get one approval request. Everything your AI submitted together arrives as one request, in To decide and as a link, with one clock. The email that carries the link shows a short summary of the record (a title, name, or text cut to about 120 characters), not the full record. If your workspace has Slack connected, Slack gets a notice instead: it names who has a request waiting and what kind, and links to To decide. It never carries the approval link.

4. You decide on each item. You read it, tick that you have reviewed it, and approve it, send it back or reject it.

5. The request lands on your last decision. Only then do the items you approved go on your record, each as its own entry.

Each entry records who approved it, when, the words they ticked, and what the safety check found, in a tamper-evident audit log. That entry cannot be quietly edited or removed: the log would show it.

For the details of each action, see Deciding on an approval request, item by item.

What is saved without an approval

• What you enter yourself. Your account, workspace, plan, billing and legal consent at signup, and the settings you change in the dashboard (approval channels, your framework licence statement, workspace settings), including inviting people and setting roles. An auditor invite needs an end date and a reason. You are the person acting, so there is no second click.
• Your company profile. You can fill it in on the dashboard, or your AI can save it for you when you describe your company. Either way it is saved directly and recorded on the audit log. It shapes later checks, so review what your AI saved.
• Drafts and working notes. What your AI proposes, the packages it opens, and any draft sent back for revision are recorded as drafts. They are not part of your approved record until someone approves them.
• Kanonik's own records. Check results, approval and delivery events, and billing events are written by Kanonik itself.

Two more things to know

• Some records get rule checks but no model check. Decisions only a person can make (exceptions, risk acceptances, remediations), records of who is asking you to prove something, and framework or questionnaire choices still need your approval. Kanonik checks them with fixed rules, not with a model.
• In a one-person workspace, you approve your own AI's work. Where a record needs a second person and you are the only member, the record marks your approval as a named exception. Other records are approved as usual.

Why saying "yes" in the chat is not enough

You might tell your AI "yes, go ahead" in the conversation. That is not approval. A reply in a chat is easy to lose, easy to dispute, and tied to a session no one else can see.

Your decision is different. It is an act you take outside the conversation, in Kanonik, on the request made for that work, and it is written into your record as the moment of authorization. Your AI cannot approve anything for you. The approval is a recorded act, not a line in a transcript, and that is what an auditor can rely on.

When a second person is involved

In a workspace with two or more people who can approve, some records need a second person. Policies, procedures, controls, decisions on what applies to you, exceptions and risk acceptances are the usual ones. For those, your click accepts what your AI prepared, and a colleague seals it later from their To decide. The record shows both of you.

If you are the only person who can approve in your workspace, you seal those records yourself after ticking a second box, and each one records a named exception saying so. The record never implies a second person looked when none did.

Links that are used or run out

A request's link works for up to an hour from when it was created. Your ticks and decisions are saved as you go. If you need longer, add an hour in To decide, a limited number of times. Adding an hour replaces the link: the emailed link stops working and you carry on in To decide.

If the time runs out before your last decision, nothing was lost and nothing was recorded. Reopen the request from To decide: it gets a fresh hour, keeps your work, and does not run the safety check again.

Once a request has landed, its link is used. Opening it again tells you the link is no longer usable; what it decided is in your Audit log.

The rules are the same for every workspace

In Settings, How approvals work is a read-only page that states the rules. Nothing on it can be changed, and every workspace gets the same values:

• A link stays open for 1 hour.
• You can add time twice in a sitting. Each addition is recorded. After that, reopen the request from the list: your work is saved.
• Up to 20 records or 10 governance items share one link and one clock. A larger run is not accepted as one group; your AI has to send it in smaller groups. A set of saved answers and the new answers it brings can share one group of up to 100 items. A larger one is refused, and your AI has to send it again in smaller groups: the new answers first, up to 20 per group, then the set.
• Links are sent by email to the person who decides, once their sign-in address is verified. If Slack is connected, it gets a notice that a request is waiting, never the link.
• Risk acceptances, exceptions, applicability exclusions and control retirements are always opened and decided one by one. A different person seals them whenever your workspace has another person who is eligible to. If nobody else is eligible (for example, you are the only member), you seal them yourself after a second acknowledgement, and each one is recorded as a named exception.

Your role

Your job is to review and decide. You do not hand-edit the policies, controls, or other compliance objects yourself. That keeps the record clean: every entry traces back to a draft the AI prepared, a check that ran, and a decision a named person made.

If an item is wrong, you do not fix it in place. You send it back with your reason, and your AI prepares a new version that is checked again and comes back to you. If you do not want it at all, you reject it, and it never lands. The record only ever holds versions someone approved.

More help

Browse every article in the Help center. For anything else, email [email protected] and a person who works on the product answers.