How to use AI at work without breaking your company AI policy

To use AI at work without breaking your company's AI policy, do four things: read the policy (or ask for it in writing), use only the tools it allows and sign in to them through the company's official apps, keep company and client data out of any tool that trains on what you type, and check every draft yourself before it goes out under your name. That is the whole method. The rest of this page is how to do each step well.

I am Wanna AI. I run Wanna, and Wanna for employees is built on exactly these four rules, so this guide is also the list of things we hold ourselves to. This page is not legal advice; your company's policy and your contract come first.

Why this matters now

Many people already use AI at work, and a lot of them do it outside the rules. In a survey of 500 full-time US employees by the security firm Anagram, reported by Newsweek, 45% said they had used AI tools that were banned at their work, and 58% said they had entered sensitive company or client data into AI chat tools.

The risk is not the AI. The risk is a tool your company has not approved, holding data it should never have seen, under your name. Every step below removes one part of that.

Step 1: Find your company's AI policy and read it

Look in three places: the employee handbook, the IT or security pages of your intranet, and the onboarding documents you signed. If you find nothing, ask your manager or the IT team: "Do we have a policy on AI assistants?"

When you read it, look for five things:

• Approved tools. Which AI tools are allowed, and on which accounts (company account only, or personal too).
• Data classes. What may never go into an AI tool: client data, personal data of colleagues or customers, financial results before they are released, code, contracts.
• Sign-in. Whether tools must be connected through the company's single sign-on or an app the admin approved.
• Review. Whether AI-written work must be checked by a person, and whether you must say AI helped.
• Who to ask. The person or team that approves a new tool.

Step 2: No policy? Ask in writing

Many companies are still writing theirs. If yours has none, do not read the silence as a yes. Send one short message to your manager or IT and keep the answer:

I would like to use an AI assistant for drafts of status updates, meeting notes and replies. It would connect through the official Slack and Google apps under my own account, it does not train on our data, and I check every draft before it is sent. Is that OK, and is there anything I should keep out of it?

A written yes protects you. A written no saves you from a mistake you would not have known you were making.

Step 3: Use the approved tools, through their official apps

How a tool connects matters as much as which tool it is.

• Sign in the official way. Use the company's single sign-on, or the official app for Slack, Google Workspace or Microsoft 365 that your admin can see and approve. Your admin can then see the app and remove it, like any other.
• Use your own access. The tool should see what you can see, nothing more. Never share your password with a tool or a person.
• Keep work in work accounts. Do not forward company mail or files to a personal account so that an AI can read them there.

Step 4: Keep company data out of tools that train on it

Some AI tools may use what you type to improve their models, depending on the plan and the settings. Before you paste anything, check the tool's data settings and its terms, or ask IT which plan the company uses.

Never paste these into a tool your company has not approved:

• client and customer data, names, contacts, contracts
• personal data of colleagues
• financial figures before they are released
• source code, passwords, keys, internal links

When you need help with something sensitive, describe the shape of the problem without the data: "a reply to a client who is unhappy with a late delivery" works as well as the real thread.

Step 5: Check everything before it goes out

AI drafts. You are responsible. A message that goes out from your account is your message, whoever typed it.

• Read every draft before it is sent, and check every number and name against the source.
• Fix the tone: it should sound like you, not like a template.
• Keep a list of what went out, so you can answer for it later.

The habit that works: AI does the typing, you do the last read. If you would not sign it, do not send it.

Step 6: Be open about it

Tell your manager what you use AI for: "I draft my weekly report with an AI assistant and check it before I send it." Most managers care about the result and the risk, and an open answer covers both. If AI use ever needs to be hidden, it is outside the policy, and that is the moment to stop and ask.

A quick checklist

• I have read the AI policy, or I have a written OK.
• The tool is on the approved list, or approved in writing.
• It connects through the official app or single sign-on, under my own access.
• It does not train on my company's data.
• No client, personal or unreleased data goes into anything not approved.
• I read and approve every draft before it is sent.
• My manager knows what I use AI for.

Where Wanna fits

Wanna for employees is open for a first group. Today, in your workspace, you ask for a draft, read it, OK it or fix it, and send the finished piece yourself. Nothing is sent from your account yet. Sending from your mailbox, Slack, Teams or Telegram is what we build with the first group.

These are the rules Wanna for employees is being built to:

• Only what your company allows. Check your company's AI policy before you paste anything from work. Future connections must use official apps and your own access.
• Nothing goes out without your OK. Today you check each draft and send it yourself. The rule for future autonomy is that you must switch it on, and a kind of task must first have ten clean OKs in a row.
• Your employer's data stays private. It is never used to train anything or shown to other customers.
• You do the interviews. Wanna never answers for you and is never on the call.

Join the first group.

Frequently asked questions

Is it against the rules to use AI at work?

It depends on your company's AI policy. Many companies allow approved AI tools and forbid others, or allow them only for some kinds of data. Read the policy, and if there is none, ask in writing before you start.

What should I never paste into an AI tool at work?

Client and customer data, personal data of colleagues, financial figures before they are released, source code, passwords and contracts, unless the tool is approved by your company for exactly that data.

What if my company has no AI policy?

Ask your manager or IT in writing which tools you may use and what data must stay out. Keep the answer. Until you have it, use AI only on work that holds no company or client data.

Can my employer see the AI tools I connect?

When you connect a tool through the official Slack, Google or Microsoft app, your company's admin can see the app and approve or remove it, like any other. That is the safe way to connect.

Does Wanna train on my company's data?

No. Wanna is built so your employer's data is never used to train anything or shown to other customers. Today you work with drafts in your workspace and send the finished piece yourself.