Human Oversight Statement

This statement explains how Wizard Application™ (the “Service”) keeps a person in control of decisions and actions that involve artificial intelligence.

Confirm before act

Wizard AI does not act on its own, with one exception that an administrator has to choose deliberately: an AI employee set to reply automatically, described below. Everywhere else, where it can carry out a task for you, such as creating a draft record, updating an item, or starting a timer, it first describes what it will do and waits for your confirmation. Anything it creates is a draft or a normal, reversible change you can edit or undo before it takes effect.

What this looks like in practice

For example, if you ask the assistant to log a new candidate, it will show you the details it plans to save, such as name, role, and source, before anything is written to your account. You can edit any field, cancel the action entirely, or confirm it. The assistant behaves the same way for other write actions, such as updating a task or starting a timer: propose, show, wait for you.

AI employees

An account administrator can add an AI employee: an AI assistant that holds a role, a department and a reporting line, and answers help desk tickets from the knowledge base of that workspace. It cannot sign in, and it has no password.

The administrator chooses whether it drafts a reply and waits for a person to accept it, or replies automatically. Automatic replies are the one case in which AI generated content reaches a third party without a person reviewing it first, and they happen only where an administrator has deliberately turned them on for that AI employee.

Where a reply is held for approval, it is sent to the people the AI employee reports to. They can edit it before sending, or discard it, and they can see which knowledge base articles it was drawn from. Only those people can accept it, and removing someone from the reporting line withdraws that ability immediately.

Every AI employee can be paused with a single switch, which stops it acting on anything while keeping its settings, and can be capped on how many replies it writes in a day and how much it costs in a month. An AI employee can never be an administrator, can never hold permissions over users, settings, the password vault, data deletion or anything that spends money, and can never be another person’s manager. Those limits are enforced whenever it acts, not only when its role is chosen.

People make the decisions

AI features are designed to support human judgement, not substitute for it. This applies in particular to recruitment features such as candidate screening, candidate-to-role matching, and CV parsing, available on our Professional and Enterprise plans: these produce suggestions, scores, and summaries for your team to review, but the hiring decision itself is always made by a person, in line with the Equality Act 2010 and other applicable employment legislation. See our AI Limitations & Accuracy Notice for more on why this matters.

Administrator controls

Account administrators can turn AI features off for the whole account, and can independently enable or disable individual capabilities such as CV parsing, candidate matching, and AI-assisted help desk replies. These controls apply account-wide rather than per user. Whatever action the assistant proposes is also bounded by the same role-based permissions as the person requesting it, so it can never do anything that person could not already do by hand, and every write action still requires explicit confirmation regardless of these settings.

Audit trail

Once you confirm an AI-proposed action, it is recorded in the same audit trail as a change made manually, so there is a clear record of what happened and who approved it.

Escalation and review

If an AI-assisted action or suggestion does not look right, you can always ignore, edit, or undo it. If you believe a safeguard has been bypassed or an action was taken without proper confirmation, please report it using our AI Content Reporting Policy.

Related policies

This statement should be read together with our AI Transparency page, our AI Limitations & Accuracy Notice, and our Acceptable Use Policy.

Contact us

If you have questions about how we keep humans in control of AI-assisted actions, contact us at support@wizardapplication.com.