AI Is Banned in Your Organization. What You Can Do Right Now
When AI is banned or restricted at work: three types of “no,” how to handle today’s task, and how to prepare a future AI pilot without breaking the rules.
Imagine you urgently need an analytical memo for tomorrow morning’s meeting. AI could help you put it together in a fraction of the time, but in your organization AI is banned or restricted.
In that situation, you will almost certainly have to prepare today’s analysis the traditional way.
Even on this very task, however, you can identify which parts of the work you could delegate to an AI model in the future — so you can eventually offload time-consuming routine and devote more energy to high-value strategic work.
Important: the decision on whether and how to use AI belongs to your employer; bypassing a ban is never acceptable. Using a personal phone or a home laptop does not override your employer’s prohibition against using AI for work purposes or uploading internal data.
There Are Three Kinds of “Not Allowed”
First of all, read your organization’s information security policy. Look for answers to two questions: is there an approved AI tool for your task, and what data are you permitted to use in it? If the policy doesn’t say, ask your information security (or cybersecurity) team. After all, nobody wants to end up in the headlines like the Samsung engineers who pasted proprietary data into ChatGPT — see the details in our Telegram post.
Let’s break down the three kinds of “not allowed”:
- External AI tools are allowed, but uploading internal documents and data is prohibited. If your organization’s rules permit general queries, you can ask an AI model to generate a generic outline for an analytical memo without any company data — and then fill it in manually inside your corporate environment.
- There is no approved AI tool. You prepare the memo without AI.
- There are no rules on AI use yet. The absence of rules is neither permission nor a permanent ban. Until you receive official clarification, prepare the memo without AI, just as in the second case.
1. External AI Is Allowed, but Work Data Cannot Be Uploaded
Here you can ask AI to design the general structure of the document while doing all work with internal sources yourself. (If your organization already has a solid memo template for this situation, just use it — there is no point prompting a neural network to reinvent an existing template.)
Take an imaginary example: ahead of a morning meeting, your manager asks for a brief memo explaining how far your department is behind plan, what is known about the causes, and what actions you propose. You have three internal sources: 1) your department’s report containing metrics and staff explanations, 2) an adjacent department’s spreadsheet with related execution figures, and 3) the minutes from last week’s meeting listing prior decisions. None of these documents contains a ready-made answer to the manager’s question.
If you don’t have a suitable template and general AI queries are permitted, you can prompt the AI like this:
Propose an outline for an analytical memo on plan execution. Length: one page. Reader: department manager. It needs to cover plan, actuals, variances, their causes, and a proposed course of action. Specify what belongs in each section. Do not invent data or conclusions.
Notice that the prompt contains only generic structural requirements. None of the reports, departmental figures, or meeting decisions are included. Simply deleting the department name from a work report before uploading it to an external AI does not make it safe to share: internal company information is still inside.
For this memo, a clean structure looks like this:
- Where we stand against the plan. Plan, actuals, size of the gap, and the source of the numbers.
- What is known about the causes. Confirmed causes versus assumptions that still need to be verified.
- What is proposed. The recommended action, its rationale, and the decision required from the manager.
Compare the AI’s outline against the goal of your meeting: does it cover every required question, and are there any redundant sections? Then copy the chosen outline into a working document inside your organization’s perimeter, fill it in using your internal sources, verify the numbers, and substantiate your proposed action.
AI helped structure the memo, while you prepared and verified its content yourself. Do not paste the completed document back into the external AI for polishing: it now contains internal data whose transmission is prohibited.
2. There Is No Approved AI Tool
Prepare today’s memo the usual way. If you want to propose an AI trial (pilot) to your employer in the future, start from a concrete task: which specific part of the workflow you expect to simplify, and how you will know whether it worked.
In our example, you might formulate a testable hypothesis: AI compiles a rough summary of metrics and explanations across the three source documents, while the employee verifies it against the originals and formulates the recommendation for the manager. Running such a trial will require separate approval for both the tool and the data class. For now, you can prepare the baseline.
To evaluate the benefit of AI later, you first need to know how much effort the same work takes today. The next time you prepare a memo like this, log three things in a working note:
- What deliverable is needed and from which sources. For example: a one-page plan-execution brief based on the department report, the adjacent team’s spreadsheet, and prior meeting minutes.
- Where the time actually goes: gathering inputs, drafting text, reviewing, and revising. Include the time spent by colleagues who review the memo, and log time spent waiting for data or approvals separately.
- What had to be fixed and why. For example: a number from the wrong period slipped into the summary, or a proposed action lacked supporting grounds.
Repeat this log across similar tasks over two to three weeks. That will reveal which stage consistently consumes time versus what was a one-off hiccup. You may discover that most of the elapsed time is spent waiting for figures from another department. In that case, speeding up draft writing alone won’t fix the delay — the bottleneck sits upstream with an adjacent team.
Armed with these observations, you can discuss a concrete pilot with your manager: which step to delegate to AI, which documents will be involved, and how the outcome will be evaluated. If the organization approves the trial, compare tasks of similar complexity, counting the active working time of all participants (including review and corrections) as well as waiting time for inputs and sign-offs. Remember: a fast rough draft does not automatically mean the finished memo reaches your manager any sooner.
3. There Are No Rules on AI Use Yet
Formal rules often lag behind everyday practice. Employees may already be experimenting with AI while the organization has not yet assigned an owner or time to define which tasks and data classes are permissible. In that situation, leadership may impose a temporary blanket ban simply to halt unmanaged use until clear conditions are set.
Moving to approved use requires defining permissible tasks, tools, data boundaries, and accountability for the output. A concrete proposal gives the organization something specific to review, even though it does not guarantee approval. First, ask for an official ruling on your intended use case — and until you receive an answer, prepare your memo without AI.
Contact your information security team or the person responsible for AI governance in your organization. Instead of asking a vague “Are we allowed to use AI?”, describe the exact action you want to take. For our memo example, the request might read:
I would like to use an external AI tool to generate a generic outline for an analytical memo. I do not plan to enter any internal documents, figures, or company details. Is this mode of use permitted? If so, which service may be used and under what conditions?
Send the question through official work channels and keep the reply on record:
- If general queries without internal data are approved, proceed as in Case 1.
- If use is prohibited or the reply remains inconclusive, continue working without AI; you can still log your baseline metrics on upcoming tasks to prepare a future pilot proposal.
And if your organization deploys its own on-premise or enterprise-approved AI environment, even better.
What to Do on Your Very Next Task
- External AI is allowed, internal data cannot be shared: if needed, generate a generic document outline in AI, then populate and verify it inside your corporate environment.
- No approved tool exists: complete the task without AI. To prepare a future pilot proposal, log the required deliverable, sources, time spent per stage, and reasons for revisions.
- No rules exist yet: request official clarification for your specific use case, and work without AI until you receive an answer.