How to ask IT to unblock an AI tool at work, step by step
To get an AI tool approved at work, send a short, specific request to the right team: name the tool, one work purpose, what data will and will not go in, and what the vendor says about data. Offer limits such as a small pilot. If the answer is no, ask what would change it and accept the decision.
If ChatGPT or another AI chat is blocked at work, a clear request is the route that can actually lead to a yes. Reviewers in IT and security approve what they can understand quickly and defend later. A good request does most of that work for them, and it shows you took the company's rules seriously.
Read next
On this page
Step 1: find the right person and read the policy
Start with the company's AI or acceptable-use policy. It is often on the intranet, and it may already list approved tools and explain how to ask for others. Then find the channel. That is usually the IT service desk or a ticket portal. Some companies send new software and web access requests to the security team instead. If you are unsure, ask your line manager. They can point you to the right queue and may need to support the request anyway. One request in the right place is handled faster than five messages sent to different people.
Step 2: name one concrete work purpose
"Please unblock ChatGPT" is hard to approve because it covers everything. A reviewer cannot judge the risk of every possible use. A single purpose they can picture works better: drafting first versions of routine emails, summarizing public articles or published reports, or explaining a piece of code that contains no secrets and no customer data. Pick one. Say who does the task and how often. You can ask for more uses later, once the first one is running. A narrow request is easier to approve, to limit and to check.
Step 3: bring the facts a reviewer needs
Reviewers usually do not say no because the idea is bad. They say no, or nothing, because they must chase the missing details. Give them these up front. The tool's name and web address. What data will be entered and what will never be entered. Who will use it. Whether an account is needed. A link to the vendor's published data and privacy terms. Whether a company-approved alternative already exists, and why it does not cover your task. With these answers your request becomes a decision instead of a research project.
Step 4: a message you can copy
Subject: Request to allow [tool name] for [one specific task] Hello [name / team], I would like to request access to [tool name and URL] for [one concrete purpose, for example: drafting first versions of routine customer emails]. At the moment it is blocked on the company network. How I would use it: [who and how often, e.g. me and two colleagues, a few times a week]. Data: I would enter only [e.g. public information and my own draft wording]. I would not enter client data, personal data, source code, credentials or anything marked confidential. Account: [no account needed / a work email account would be needed]. Vendor terms: the provider's data and privacy terms are here: [link]. [If you have read them: one line on what they say about stored data and model training.] Alternatives: [we have no approved tool for this task / the approved tool does not cover this because ...]. I am happy to start with a limited pilot: [one team, no confidential data, review after 4 weeks]. I will follow the company AI policy and check all output before using it. Could you tell me whether this is possible, what else you need from me, and roughly how long the review takes? Thank you for your time. [Your name, role, team, contact]
Step 5: offer limits that make yes easy
A limited yes is much safer for a reviewer than a company-wide one. Offer a pilot for one team, for a fixed period, with a review date. Promise that no confidential, client or personal data will be entered. Offer to name one person who owns the pilot, and to report back after it on where the tool helped and where it did not. Add that a human will check every output before it is used. Limits do not weaken your request. They show you understand why the block exists and that you are helping to manage the risk.
If the answer is no, or you hear nothing
Reviews can take days or weeks. Follow up once, politely, after about a week. If the answer is no, ask what would make the request approvable, such as a vendor agreement or a business case. Ask which approved alternative you can use now. Then accept the decision. Do not work around the block. Going around it, by any route, can breach company policy and put your job at risk, even if your intent is harmless. The same method works for a library, coworking space or hotel network: ask the administrator, name the tool and the purpose, and accept the answer.
Straight answers
How long does an unblock request take?
It ranges from a day to several weeks, depending on the company and the tool. Ask the person handling it for the usual time when you send the request, and follow up once after a week.
Will I get in trouble for asking?
Asking through the proper channel is the correct route, and IT teams generally prefer a request to a hidden workaround. Keep it factual and polite, and never include confidential data in the request itself.
Who decides whether an AI tool is allowed?
Usually IT or the security team decides, sometimes with legal, data protection or a department head. Your manager can support the request but often cannot approve it alone.
ChatGPT is blocked at work. What should I do?
Read the AI policy, check whether an approved tool exists, and use that. If none exists, send IT a specific request with one purpose, the data involved and the vendor's terms.
What if IT says no?
Ask what would change the answer and which approved option you can use now. Keep the request on file for later. Do not look for a way around the block, as that can breach policy.
For general questions with no work information, this site offers free chat with no account, no sign-up, no card and no daily message limit. Use it only where your policy allows.
Open unblocked chat