AI at Work: What Managers Should Tell Employees

Managers should tell employees plainly what AI tools are approved, what work they may and may not use them for, what information must never be entered, and who checks the results. They should also explain how AI may affect roles and decisions, invite questions, and update the guidance as tools and workplace practices change.

Start with a clear, honest explanation

“AI” covers many different systems, from tools that draft text or summarize documents to software that classifies information or makes predictions. A useful workplace message names the specific tools and tasks in scope rather than implying that every AI system works the same way. Explain whether a tool is optional or required, who provides it, and where employees can find current instructions.

Do not describe AI as either a flawless expert or a replacement for all human judgment. These systems generate outputs from patterns in data and can produce convincing but incorrect material. Their capabilities vary by tool, configuration, and task. Employees need enough context to decide when an output is useful, when it needs checking, and when the task should stay with a person.

Managers should also be candid about uncertainty. If leaders have not decided whether a tool will change a process or job, say so, describe what is known, and commit to sharing updates. Unsupported promises—such as saying AI will never affect staffing or evaluation—can damage trust if circumstances later change.

Set practical rules for use and data

Give staff specific examples of permitted and prohibited uses. For example, an approved tool might help create a first draft of a routine internal announcement, while a human remains responsible for checking facts, tone, and audience. A tool that is not approved should not be used for confidential client work merely because it is easy to access.

Make the data rule concrete. Unless the organization has explicitly approved a tool and use case for that information, employees should not enter confidential business material, personal information, customer or employee records, credentials, or unpublished plans. Explain where to check whether a tool is approved and whom to ask when a task involves sensitive material. A tool’s privacy claims or familiar interface are not, by themselves, proof that it is appropriate for workplace data.

Written guidance should cover at least these points:

• Which tools and workplace accounts employees may use, and for what tasks.
• What information may be entered, what must be excluded, and how to handle data they are unsure about.
• Whether AI-generated work must be labeled, disclosed, or recorded, and in what circumstances.
• Who reviews outputs before they are shared, acted on, or used to make a decision.
• How employees report a mistake, unexpected disclosure, or harmful result.

Explain checking, accountability, and limits

AI output is a draft or suggestion, not evidence that something is true. A generated summary can omit a qualification; a suggested answer can cite a source that does not support it; and a calculation or translation can be wrong. For consequential work, employees should verify important claims against reliable source material and follow the same professional review process they would use for other work.

Be clear about responsibility: the person or team using an output remains accountable for the work they submit or act on, subject to the organization’s usual roles and review procedures. “The AI said so” is not a substitute for checking. At the same time, managers should not make employees solely responsible for risks created by unclear rules, unsuitable tools, or inadequate training. Leaders need to provide safe processes and a way to raise concerns without discouraging honest reporting.

Some tasks deserve stricter controls or should not be delegated to AI at all under an organization’s policy. Examples may include decisions about a person’s employment, access to essential services, or other high-impact matters. The appropriate boundary depends on the tool, the data, the task, and applicable organizational requirements. Managers should tell employees to escalate uncertain or high-stakes uses instead of improvising.

Discuss jobs, skills, and employee concerns

Employees may reasonably wonder whether AI will change their duties, how their performance will be assessed, or whether AI use is expected to increase their workload. Address these questions directly. Distinguish between what is already happening, what is being explored, and what has not been decided. When a process changes, explain the reason, the expected impact on daily work, and where employees can get help.

Do not assume that faster drafting means less work overall. Staff may need time to learn a tool, check its output, correct mistakes, and handle tasks that cannot be automated. Managers should make room for training and feedback, and recognize the value of careful review rather than rewarding speed alone.

Useful steps include:

• Ask employees which repetitive tasks are frustrating and where errors would be costly.
• Offer role-specific instruction, including examples of acceptable and unacceptable use.
• Make it easy to report a bad output or confusing rule without blame.
• Review whether a new workflow shifts hidden checking work onto employees or excludes people who need different ways to complete tasks.
• Explain how performance measures will treat AI-assisted work before applying them.

Make communication ongoing, not a one-time announcement

A launch email is not enough. Tools, settings, and workplace practices can change, so assign an owner for guidance and tell staff where the current version lives. Managers should coordinate messages with the organization’s relevant technology, privacy, security, human resources, and legal teams rather than making promises or inventing rules on the spot.

Keep a channel open for questions and publish answers to recurring issues. If a tool produces a harmful or materially wrong result, pause the affected use, preserve the information needed to investigate in line with company procedures, and report it through the stated route. Do not encourage employees to conceal errors or paste sensitive material into a second tool to “fix” the first result.

Review the guidance after meaningful changes, such as adding a new tool, expanding its use to a new task, or discovering a recurring failure. Ask whether employees understand the rules and whether the process is working in practice. The goal is not simply to increase adoption; it is to use tools where they help without weakening accuracy, privacy, fairness, or accountability.

A message managers can adapt

A concise announcement could say: “We are allowing the approved tools for the tasks listed in our guidance. Do not enter confidential or personal information unless the specific tool and use are approved for it. Check AI-generated facts and follow the normal review process before sharing or acting on the output. Ask your manager or the designated support team when a use is unclear, and report errors or concerns through the stated channel. We will share updates as we learn more; here is what we know today and what is still undecided.”

That wording is only a starting point. It should be adapted to the organization’s actual tools, data rules, and decision processes. The most credible communication is specific enough to guide daily work and honest enough to distinguish firm rules from open questions.

Frequently Asked Questions

Should employees be allowed to use public AI tools for work?
Only when the organization’s policy permits the specific tool and task. Employees should not enter confidential or personal information unless that use has been explicitly approved.

Can AI-generated work be trusted without review?
No. AI can produce plausible errors or omit important details. Check claims against reliable sources, especially before using output in consequential work.

Should employees disclose that they used AI?
Managers should state the organization’s disclosure rules. Disclosure may depend on the task, audience, and policy; employees should not have to guess whether a label is required.

Will AI replace employees’ jobs?
There is no universal answer. Effects depend on the organization and work involved. Managers should distinguish confirmed plans from possibilities and communicate changes when decisions are made.

What should an employee do if AI gives a wrong or harmful answer?
Do not rely on or pass along the output. Follow the normal correction process and report the issue through the organization’s designated channel, especially if sensitive information or a consequential decision is involved.

Related

• Power Analysis Calculator for Statistical Study Design
• AI Toxicity Scores: How They Work and Their Limits
• AI Support Agents Without RAG: Budget and Trade-Offs