AI Transparency

Artificial intelligence is part of how Wizard AI works. This page explains, in plain language, where we use AI, how we keep people in control of it, what happens to your data, and the governance we apply to keep our use of AI responsible and accountable.

Our approach to responsible AI

We build AI features to help your teams work faster, not to make consequential decisions on their own. Our use of AI is guided by a small set of principles:

• Human first. AI assists people. It does not replace human judgement on decisions that affect customers, employees, or your business.
• Transparency. Where a feature is powered by AI, we make that clear, and we are honest about what it can and cannot do.
• Privacy by design. AI features respect the same data protection commitments as the rest of the platform.
• Accountability. We maintain internal ownership, review, and oversight of every AI capability we ship.

How we use AI in the product

AI is used to support specific, well defined tasks inside the platform, such as:

• Drafting and summarising text — suggesting replies, summarising long threads, and generating first drafts of documents.
• Smarter search and suggestions, including semantic search that finds people or content by meaning rather than exact keywords (for example, matching candidates to a role).
• Generating an image from a text description, when you ask for one.
• An in-app assistant you can chat with — by typing, or by voice in our mobile app — to find information and carry out tasks, always with your confirmation.
• Surfacing relevant insights and highlighting items that may need your attention.

AI suggestions are always editable. You decide whether to accept, change, or ignore them before anything is saved or sent.

AI employees

An AI employee is an AI assistant that an account administrator adds to a workspace as a named team member. It holds a role, a department and a reporting line, it occupies a team member seat in the same way a person does, and it appears under its own name where it takes part in a conversation.

An AI employee cannot sign in. It has no password and no login of any kind, and it works only inside the workspace it belongs to.

Its current purpose is to answer help desk tickets using the knowledge base articles held in that workspace. An administrator chooses how it behaves:

• Draft and ask first. It writes a reply and notifies the people it reports to. Nothing is sent to the person who raised the ticket until one of them reviews it and accepts it. The draft can be edited before it is sent, and the articles it drew from are shown alongside it.
• Reply automatically. It answers new tickets itself, without a person reviewing each reply first. This is the one place in the platform where AI generated content can reach a third party without prior human review, and it applies only when an administrator has deliberately chosen it for that AI employee.

In either mode, a reply is published under the name of the AI employee, and the email address shown on it is one the account chooses. An AI employee stands back from a ticket that a person has already replied to.

Administrators keep the following controls over every AI employee:

• A pause switch that stops it acting on anything, keeping its configuration intact.
• A limit on how many replies it may write in a day, and on how much it may cost in a calendar month.
• Its reporting line, which decides who is able to approve the replies it drafts. Removing a person from that list withdraws their ability immediately.

An AI employee cannot be an administrator, and it cannot be given a role carrying user administration, account settings, the password vault, data deletion, or any capability that spends money. These restrictions are applied every time it acts, not only when its role is assigned, so a later change to a role cannot widen what it is able to do. An AI employee also cannot be another person’s manager.

The account remains responsible for the content its AI employees send, in the same way it is responsible for what its people send. We recommend reviewing replies before enabling automatic mode, and keeping the knowledge base accurate, since that is what the replies are drawn from.

Human oversight and control

AI features are designed so that a person stays in control. With one exception, described under AI employees below, Wizard AI never acts on its own: where it can carry out a task for you in the app — such as creating a draft record, updating an item, or starting a timer — it first describes what it will do and waits for you to confirm. Anything it creates is a draft or a normal, reversible change you can edit or undo. Account administrators can manage whether AI features are available, and whether the assistant may carry out actions at all.

Your data and AI

Protecting your data is central to how we design AI features:

• We do not use your private customer content to train foundation models, and we do not sell your data.
• Your content is processed only to deliver the feature you have asked for, and is handled under the same terms as the rest of your data on the platform.
• Access to data used by AI features is limited and logged, consistent with our wider security practices.

For full detail on how we collect, use, and protect personal data, please read our Privacy Policy.

To answer you, the assistant may send our model provider the text of your request, relevant records it looks up to fulfil it, any file or image you attach, and a snapshot of the page you are viewing at the time. It can only access data you already have permission to see in the platform, we send no more than is needed to provide the feature, and your content is never used to train AI models.

Memory and personalisation

To give more relevant help, Wizard AI can remember a small number of facts you tell it — such as how you prefer it to respond, your role or team, or what you are currently working on. It uses these to personalise its assistance across conversations.

• It stores short notes you share, not a copy of your conversations, and never passwords or sensitive credentials.
• What it remembers is private to your own user and is treated as your data under the same terms as everything else on the platform.
• It is never used to train AI models.
• You are always in control: you can view everything Wizard AI remembers about you, delete any individual item, or clear it all at any time from the assistant’s memory panel.

Voice input

In our mobile app you can speak to the assistant instead of typing. When you do, your device’s built-in speech recognition, provided by your phone’s operating system, converts your speech to text, which is then handled exactly like a typed message. Voice input is optional and is only active while you are using the microphone button.

Third party AI providers

By default, some AI features are powered by OpenAI, a trusted third party model provider, using wallet credits on your account. When we send content to OpenAI to generate a response, it is processed under contractual terms that protect the onfidentiality of your data and restrict its use to providing the service. OpenAI does not use content sent through our service to train its models, acts as our processor, and is covered by our data protection arrangements.

You can also bring your own AI provider. Instead of using our default OpenAI connection, you may configure the platform to use your own AI account and API key. Wizard AI works with a range of supported providers, including:

• OpenAI — our default, on wallet credits
• Anthropic — Claude models
• Google Gemini — Google models
• Azure OpenAI — enterprise Azure hosting
• Mistral — open models
• Groq — fast inference
• OpenRouter — hundreds of models through a single key
• Custom or self-hosted — any OpenAI-compatible endpoint, such as DeepSeek, Together, xAI, or your own hosted models

When you connect your own AI provider, please note the following:

• Your provider, your terms: Content processed through a provider you connect is handled under your agreement with that provider, including their privacy, data retention, and model training policies. Our data protection arrangements with OpenAI do not extend to third party providers you configure yourself.
• Your responsibility: You are responsible for reviewing and accepting that provider's terms, for the security of the credentials you supply, and for ensuring the provider is an appropriate choice for the data you process.
• Credentials: Any API keys you provide are stored securely and used only to route yyou have chosen.
• Your control: You can change or remove your configured provider at any time, after which affected AI features will revert to our default OpenAI connection or become unavailable, depending on your settings.

Accuracy and limitations

AI systems can make mistakes. Generated text may occasionally be inaccurate, incomplete, or out of date, and it should not be relied upon as professional, legal, or financial advice. We ask that you review AI generated content before you act on it, especially for anything important. We continue to test and improve our AI features, but no AI system is perfect.

Records and retention

To keep AI use secure and accountable, we keep a limited record of interactions with the assistant, such as the request made and the response given. These records are retained for a limited period and then automatically deleted, in line with our data governance practices. You can also review and delete what the assistant remembers about you at any time, as described above.

Governance and accountability

We manage our use of AI through a defined internal process, with clear ownership and regular review. We are developing these practices in line with ISO/IEC 42001, the international standard for artificial intelligence management systems, which sets out how an organisation should govern, monitor, and continually improve its use of AI. This complements the wider information security and privacy standards we work to, which you can read about on our Security page.

Your choices

You remain in control of how AI fits into your work. You can choose whether to use AI assisted features, and account administrators can manage their availability. You can also use your own AI provider and API key instead of our default OpenAI connection, see our AI integrations page for the full list of supported providers. If you have questions about how a particular feature uses AI, we are happy to explain.

Contact us

If you have a question or concern about our use of AI, or you would like to report an issue with an AI feature, please contact us at support@wizardapplication.com.