Understand threats and change

UAE survey: only 1 in 10 firms felt ready for AI tools that take actions

Can five staff habits replace company rules for using AI?

No, these five habits help staff use AI more safely, but employers still need to decide which tools staff may use and what information they may share.

Employers should explain these rules in plain language and give staff one place to ask questions. They should also decide what work files an AI tool may reach and which actions a person needs to review.

A person reviews an AI assistant on a laptop while selected work information stays inside a clear access boundary.

Some UAE employees regularly use AI tools their employers have not approved: 43% of surveyed chief information officers said this happens 1Reference 1: New IDC study identifies governance gap as AI adoption accelerates across UAE enterprises. More than half of surveyed UAE organisations had AI agents deployed or in active development, yet only one in ten considered themselves fully prepared to deploy them safely 1Reference 1: New IDC study identifies governance gap as AI adoption accelerates across UAE enterprises. Staff can start with five practical habits while employers clarify which tools, data and access are allowed.

A person reviews an AI assistant on a laptop while selected work information stays inside a clear access boundary.
A user reviews an AI assistant while keeping its access limited to selected work information. Image credit [4]

Five steps for safer AI use at work

Some AI assistants answer questions. AI agents can also connect to work files and apps and take actions, depending on the permissions they receive 3Reference 3: Apply responsible AI. Think of access like a work key: give the tool only the room it needs, not the whole building.

  1. Use only an AI service your company approves. Sign in with your work account; ask IT if you are unsure which one to use.
  2. Protect private information. Never paste passwords or access keys into an AI chat. Use customer records, personal data and confidential plans only in a tool your company has approved for that data.
  3. Limit access to the task. Do not give an AI app full access to your computer, mailbox or work drive.
  4. Check important answers and actions before they go out. Verify key facts and review any message, edit or deletion before it happens.
  5. Pause and contact IT if the tool asks for broad access, reveals data you did not provide, or proposes an unrequested action. If no approved tool fits the task, ask IT what to use instead.

Good habits need company controls

Staff habits do not replace company controls. Employers still need approved tools, clear data rules, access limits and oversight.

The Microsoft, CPX and IDC study surveyed 175 senior security and technology executives across UAE enterprises and was released on 16 October 2026 1Reference 1: New IDC study identifies governance gap as AI adoption accelerates across UAE enterprises. Only 15% of surveyed chief information security officers had mandatory authority to approve AI deployments, and just 10% of organisations set security requirements before evaluating an AI solution 1Reference 1: New IDC study identifies governance gap as AI adoption accelerates across UAE enterprises. These sponsor-backed results reflect executive responses, not a census of UAE companies.

The UK National Cyber Security Centre recommends easy ways to request tools and report concerns without blame 2Reference 2: Shadow IT guidance. A team can start by sharing this list and naming one person to approve tools, answer questions and receive reports. For wider security basics, see our small-business cyber security guide.

Sources

  1. New IDC study identifies governance gap as AI adoption accelerates across UAE enterprises

    Microsoft, CPX and IDC CPX Published 16 Oct 2026 Source language: en Accessed 21 Oct 2026
  2. Shadow IT guidance

    National Cyber Security Centre NCSC Source language: en-gb Accessed 21 Oct 2026
  3. Apply responsible AI

    Microsoft Microsoft Learn Published 14 Jul 2026 Source language: en Accessed 21 Oct 2026
  4. AI access boundaries at work (AI-generated illustration)

    Image credit: Aeptus, using OpenAI image generation License: OpenAI Terms of Use (output ownership, to the extent permitted by law)