content

Using AI at Work

Using AI at Work is a three-page intranet learning site with a fictional worked example, prompt comparison, output checks, and guidance for verifying approved tools and information before sharing.

Request pricing See it in MangoApps

Every employee gets a seat — priced per employee in Communications & Intranet, quoted with this template ready.

Rolled out to every employee at AutoZone (125,000), PetSmart (50,000+), A.S. Watson and Raley's (20,000) — and at larger retailers we are not permitted to name.

Built for: Professional services · Healthcare administration · Financial services · Higher education · Government and public services

9:41

Sites

Overview

Using AI at Work is a three-page intranet learning site for employees considering approved AI tools for bounded tasks. The Choose a useful task page sets the starting rule: select work with a source the employee can check, and verify both tool approval and information permissions before entering anything. Its sample tasks are summarizing a nonsensitive agenda, clarifying instructions, and preparing a first draft.

Work through an example follows a fictional Northstar Services project coordinator from a vague prompt to a specific request, then contrasts an AI draft that invents a checklist owner with a corrected, human-reviewed summary. A second rewrite example shows how to preserve requirements while improving clarity. The page ends with a practical sequence for checking the source, bounding the task, inspecting the output, and deciding whether it is suitable to share. Information and review provides a four-part check covering the tool, information, output, and audience, plus guidance for pausing when permission or facts are unclear.

Use this template as introductory employee guidance, a starting point for an AI-use learning hub, or a resource for a team session. Before publishing, replace or clearly retain the fictional framing, confirm internal guidance and help routes, and link to authoritative company resources. This site does not grant approval to use a tool, define your organization's data rules, automate review, or substitute for legal or policy guidance. Do not use it as a high-stakes decision procedure or as permission to enter sensitive information; direct those questions to the responsible internal owners.

Standards & compliance context

The site offers practical checks, not legal advice or a compliance determination; link to the organization's current policies and consult responsible legal, privacy, and security owners where needed.
Keep tool approval and information classification as separate checks, and preserve any scope or conditions stated in the authoritative internal guidance.
Do not use a tool's availability, another team's practice, or an AI-generated answer as evidence that a particular task or data type is permitted.
Use WCAG 2.1 AA as an accessibility target for the employee-facing pages, including meaningful image alternatives, clear headings, and readable comparison and table content.

General regulatory context for orientation only — verify current requirements with counsel or the relevant agency before relying on this template for compliance.

What's inside this template

Choose a useful task

Make AI useful. Keep judgment human.
Fictional worked example: turn a nonsensitive meeting agenda into a useful draft, catch an invented owner and decide what is ready to share.
Start with a small, checkable task
A fictional Northstar Services example for employees using approved AI tools. An approved assistant can help turn a nonsensitive meeting agenda into a concise summary or make instructions easier to follow. Choose work where you can check the result against its source, keep the task bounded, and leave decisions with the people responsible for them. Before entering information, check that both the tool and the information are permitted for your intended use. A tool being available does not approve every data use.
Good first tasks have a clear boundary
Choose work where you can recognize an accurate result. Summarize an agenda Turn a nonsensitive agenda into a short list of topics and questions to resolve. Keep the source within approved information rules. Check every point against the agenda. Clarify instructions Ask for plain-language wording while preserving the original steps and requirements. Compare the rewrite with the source. Restore any omitted requirement. Prepare a first draft Use a permitted, nonsensitive source to structure a draft that a person will review. Name the audience and format. Treat the output as a draft, not an approval.
Continue with the worked example
Follow one task from prompt to human-checked result. Work through an example Compare a vague prompt, a more specific prompt, and a summary with a fabricated owner. Check information and review Use a short decision sequence before entering information or sharing output.

Work through an example

One agenda, a better prompt, and a necessary check
Northstar Services is fictional; the agenda and outputs below are illustrative. Scenario: A project coordinator wants a concise summary of a nonsensitive planning agenda and a clearer version of a short instruction for colleagues. The agenda says the group will review the launch checklist, discuss open questions, and decide whether another review meeting is needed. It does not name an owner for the checklist. The aim is to produce a draft that saves time while keeping the source, uncertainty, and human decision visible.
Make the request specific
The improved prompt sets scope and tells the assistant not to fill gaps. Prompt Summarize this meeting agenda and make the instructions clearer. Using only the nonsensitive agenda text below, write a summary of up to three bullets covering the topics and decisions requested. Do not infer decisions, owners, or dates. If the agenda names none, say so. Then rewrite the instruction in plain language without changing its requirements. Keep both drafts separate from the source.
Check the output against what was actually provided
The flawed draft sounds confident; the source does not support its added owner. Flawed AI draft “The team will review the launch checklist. Jordan will own the checklist updates, and the group will decide whether to hold another review meeting.” The name and ownership claim were not in the agenda. The agenda names no checklist owner. A plausible detail is still an invented detail. Do not forward this draft as an agreed action. Corrected, human-reviewed summary “The group will review the launch checklist, discuss open questions, and decide whether another review meeting is needed. The agenda does not identify an owner for checklist updates.” A reviewer compared each sentence with the agenda and removed the invented name. Topics match the source. The missing owner remains explicit. The group—not the assistant—makes the meeting decision. Clearer instruction draft Source instruction: “Please review the agenda before the meeting and come ready to discuss it.” Draft: “Read the agenda before the meeting. Bring any questions or points you want the group to discuss.” The reviewer confirms that the rewrite has not added a deadline or changed the requested action. Compare the rewrite with the original. Keep requirements and meaning intact. A person decides whether the wording is suitable to share.
Use the same review sequence on your own task
If the source or result is unclear, pause and ask the relevant owner. Check the source Confirm the information is suitable for the approved tool under your company's current tool and information guidance. Set a bounded task State the audience, output format, and what the assistant must not infer or change. Inspect the draft Compare names, owners, dates, decisions, and requirements with the source. Remove or correct unsupported claims. Choose the handoff Share only a reviewed result that is appropriate for its audience. If a claim cannot be verified, return to the source owner rather than treating the AI output as evidence.
Check the information rules before trying it
Information and review See how to check tool approval, information class, and the right human review.

Information and review

Tool approval is not blanket permission
Check both the tool and the information before you enter anything. An assistant being available to employees does not mean every task or information type is approved for it. Consult your company's approved-tool register and information-classification guidance. Follow the specific scope and conditions shown there; do not infer permission from a tool's presence or from another team's use. If either the tool's approval status or the information's classification is unclear, pause. Ask the role responsible for the relevant tool or information guidance, or use the company's designated AI help route. Do not paste confidential source material into a prompt to find out whether it is allowed.
A practical check before sharing
If a check fails or is uncertain, do not proceed on assumption. Check What to verify If unclear Tool The assistant is listed as approved for this task. Ask the tool owner or designated AI help route. Information The source is permitted under its classification guidance. Ask the information owner or follow the classification guidance. Output Names, dates, owners, decisions, and requirements match the source. Correct it from the source or ask its owner. Sharing The audience can receive the reviewed content. Ask the content owner before sharing.
Common questions
A tool can assist with a draft; responsibility for using and sharing it remains with people. Can I use a tool just because it is available? No. Check the approved-tool register and the information-classification guidance for the intended task and source. Availability alone does not approve every use. What if the assistant supplies an owner or decision that is missing from my source? Treat it as unverified. Remove it or check with the person or role responsible for the source. Do not present an inferred detail as an agreed fact. How can I share a useful learning? Share a short description of the task, the prompt pattern, and a lesson such as “ask the assistant to flag missing owners.” Leave out confidential source text, personal details, and unreviewed output. Use the company's designated AI help route when a question needs an authoritative answer.

How to use this template

Clone the site and confirm the audience, content owner, approved-tool register, information-classification guidance, and designated AI help route for your organization.
Replace the fictional Northstar Services scenario only with a clearly labeled example consistent with current internal guidance, and keep sensitive or personal source material out of sample prompts.
Add working links to authoritative internal guidance pages where the template directs employees to check tool approval, information rules, or help.
Assign the learning site to its content owner and ask the relevant tool and information owners to verify their guidance and escalation routes before launch.
Publish the three pages together in the supplied navigation order, then direct employees to begin with a small task whose output they can compare with its source.
Review the pages when underlying guidance or help routes change, and correct or retire any example that no longer reflects permitted practice.

Best practices

Keep the fictional Northstar Services framing visible unless you replace it with a verified, clearly approved internal example.
Use examples with a source employees can inspect, and explicitly mark facts such as owners or dates that the source does not provide.
Link the tool and information checks to authoritative internal pages rather than implying that this learning site grants permission.
Preserve the prompt's task boundary, audience, output format, and instruction not to infer missing decisions, owners, or dates.
Tell reviewers to compare names, dates, owners, decisions, and requirements with the source before sharing a draft.
Provide a real, maintained help route for unclear approvals and identify the responsible information owner for classification questions.
Keep page headings and navigation labels aligned so employees can move from task choice to example to information review without searching.

What this template typically catches

Issues teams running this template most often surface in practice:

Employees assume that an available assistant is approved for every task or information type.

A prompt omits boundaries, allowing the assistant to infer owners, dates, decisions, or requirements.

A plausible invented detail survives review because the draft sounds confident.

A rewrite changes the meaning or drops a requirement from the original instruction.

Employees share unreviewed output or send it to an audience that should not receive it.

Guidance points to an outdated or missing tool register, classification page, or help route.

A learning site has no named content owner or review trigger, so examples and links become stale.

Common use cases

Employee AI onboarding

An intranet content owner uses the three-page sequence to orient employees to bounded tasks, demonstrate source checking, and point them to current internal guidance. The fictional scenario gives learners a concrete review exercise without implying company approval.

Project coordinator prompt practice

A project coordinator uses the agenda example to practice requesting a short summary without invented owners or decisions. The reviewer compares the result with the agenda and returns unresolved facts to the source owner.

Information governance guidance

A tool or information-governance owner adapts the Information and review page to link employees to authoritative approval and classification guidance. The page supports a pause-and-ask response when either permission is unclear.

Manager-led team learning

A manager walks a team through the before-and-after prompt and asks participants to identify the unsupported checklist owner. Team members then apply the same source, output, and sharing checks to a permitted, nonsensitive task.

Frequently asked questions

Who is this intranet site for?

It is designed for employees who use company-approved AI tools for bounded drafting, summarizing, or rewriting tasks. The pages show how to check a result against its source and keep decisions with responsible people. Tool availability does not mean every employee, task, or information type is automatically in scope.

What does the template include?

The site has three pages: Choose a useful task, Work through an example, and Information and review. It includes sample prompts, a fictional agenda scenario with an invented owner to catch, a human-reviewed correction, and checks for tool approval, information, output, and sharing. The Northstar Services example is fictional and should not be presented as your company's policy.

How often should we review or update the pages?

Review the guidance when your approved-tool register, information-classification rules, or designated help route changes. Assign a content owner to check those dependencies on the cadence your organization uses for policy and intranet content. Update the worked example if its prompt pattern no longer reflects permitted practice.

Who should own rollout and employee questions?

A suitable owner is the role responsible for employee AI guidance or intranet learning content, working with the owners of tool approval and information classification. Name your organization's actual help route before publishing; the template does not supply a real contact or approval process. Direct unclear tool questions to the designated AI help route and information questions to the relevant information owner.

Does this template make AI use legally or regulatorily compliant?

No. It teaches practical checks but does not establish legal compliance or replace company policy, regulatory advice, or a formal risk assessment. Confirm applicable requirements with your organization's legal, privacy, security, and information-governance owners. Preserve the distinction between an illustrative example and authoritative company guidance.

What is the most important pitfall the example teaches?

An AI draft can confidently add a plausible owner that the source never named. The example tells the reviewer to compare names, owners, dates, decisions, and requirements against the source, then remove unsupported claims. If a fact cannot be verified, ask the source owner rather than treating AI output as evidence.

Can we use the example with sensitive information or any AI assistant?

The example uses a nonsensitive fictional agenda and does not authorize use with sensitive or confidential material. Replace the sample only with scenarios permitted by your current tool approval and information-classification guidance. Employees should check both the tool and the information for the intended task before entering content.

What should we customize before publishing?

Replace the fictional company and scenario with a clearly labeled example that fits your approved practices, and add links only to real internal guidance pages. Confirm the approved-tool register, classification guidance, content owner, and help route exist and are current. Keep the sample's core lesson intact: bound the task, verify the output against its source, and make sharing decisions through the appropriate human review.

How does this compare with an ad-hoc AI tip sheet, and does it integrate with AI tools?

The three-page structure gives employees a path from choosing a task to reviewing an example and checking information before sharing, rather than presenting disconnected tips. It is instructional content, not an AI workflow or tool integration. It does not send prompts, check permissions automatically, or connect to a live tool register; link or connect those resources through your site's supported setup.

Related templates

Sites

Working With External Collaborators

A one-page decision guide for employees sharing work with external collaborators. Compare three a...

Sites

Useful Search and Knowledge Requests

A practical employee learning page for refining internal searches, checking whether an answer is ...

Sites

Understanding Your Total Reward Statement

This employee reference page explains how to read a total reward statement, distinguish paid amou...

Sites

Performance Review Preparation

Prepare for a performance-review conversation with prompts for agreed outcomes, observable eviden...

Sites

Performance Calibration Session

A single-page performance calibration session guide for people leaders and HR partners, with an i...

Forms

Employee Onboarding Form

Employee Onboarding Form for collecting new hire details, tax references, direct deposit, emergen...

Inspections

Abandoned Channel Hygiene Audit

Abandoned Channel Hygiene Audit template for reviewing inactive, orphaned, and duplicate communic...

Sop

New Employee Onboarding Checklist

New Employee Onboarding Checklist template for standardizing pre-boarding, first-day orientation,...

Go deeper on the topic

Related concepts

AI Governance

AI governance is the framework a company uses to decide what AI tools are allowed to do, who's accountable for their outputs, what data they're allowed to...
Digital Workplace vs Modern Intranet

A modern intranet is a specific surface — typically the home-base destination where employees get company news, find policies, and access key apps. A digital...

Ready to use this template?

Every employee gets a seat. Request pricing for Communications & Intranet and we quote into a workspace with Using AI at Work ready.

Request pricing

Rolled out to every employee at AutoZone (125,000), PetSmart (50,000+), A.S. Watson and Raley's (20,000) — and at larger retailers we are not permitted to name.

3 pages Choose a useful task Work through an example Information and review

Choose a useful task Make AI useful. Keep judgment human. Fictional worked example: turn a nonsensitive meeting agenda into a useful draft, catch an invented owner and decide what is ready to share. Start with a small, checkable task A fictional Northstar Services example for employees using approved AI tools. An approved assistant can help turn a nonsensitive meeting agenda into a concise summary or make instructions easier to follow. Choose work where you can check the result against its source, keep the task bounded, and leave decisions with the people responsible for them. Before entering information, check that both the tool and the information are permitted for your intended use. A tool being available does not approve every data use. Good first tasks have a clear boundary Choose work where you can recognize an accurate result. Summarize an agenda Turn a nonsensitive agenda into a short list of topics and questions to resolve. Keep the source within approved information rules. Check every point against the agenda. Clarify instructions Ask for plain-language wording while preserving the original steps and requirements. Compare the rewrite with the source. Restore any omitted requirement. Prepare a first draft Use a permitted, nonsensitive source to structure a draft that a person will review. Name the audience and format. Treat the output as a draft, not an approval. Continue with the worked example Follow one task from prompt to human-checked result. Work through an example Compare a vague prompt, a more specific prompt, and a summary with a fabricated owner. Check information and review Use a short decision sequence before entering information or sharing output. Work through an example One agenda, a better prompt, and a necessary check Northstar Services is fictional; the agenda and outputs below are illustrative. Scenario: A project coordinator wants a concise summary of a nonsensitive planning agenda and a clearer version of a short instruction for colleagues. The agenda says the group will review the launch checklist, discuss open questions, and decide whether another review meeting is needed. It does not name an owner for the checklist. The aim is to produce a draft that saves time while keeping the source, uncertainty, and human decision visible. Make the request specific The improved prompt sets scope and tells the assistant not to fill gaps. Prompt Summarize this meeting agenda and make the instructions clearer. Using only the nonsensitive agenda text below, write a summary of up to three bullets covering the topics and decisions requested. Do not infer decisions, owners, or dates. If the agenda names none, say so. Then rewrite the instruction in plain language without changing its requirements. Keep both drafts separate from the source. Check the output against what was actually provided The flawed draft sounds confident; the source does not support its added owner. Flawed AI draft “The team will review the launch checklist. Jordan will own the checklist updates, and the group will decide whether to hold another review meeting.” The name and ownership claim were not in the agenda. The agenda names no checklist owner. A plausible detail is still an invented detail. Do not forward this draft as an agreed action. Corrected, human-reviewed summary “The group will review the launch checklist, discuss open questions, and decide whether another review meeting is needed. The agenda does not identify an owner for checklist updates.” A reviewer compared each sentence with the agenda and removed the invented name. Topics match the source. The missing owner remains explicit. The group—not the assistant—makes the meeting decision. Clearer instruction draft Source instruction: “Please review the agenda before the meeting and come ready to discuss it.” Draft: “Read the agenda before the meeting. Bring any questions or points you want the group to discuss.” The reviewer confirms that the rewrite has not added a deadline or changed the requested action. Compare the rewrite with the original. Keep requirements and meaning intact. A person decides whether the wording is suitable to share. Use the same review sequence on your own task If the source or result is unclear, pause and ask the relevant owner. Check the source Confirm the information is suitable for the approved tool under your company's current tool and information guidance. Set a bounded task State the audience, output format, and what the assistant must not infer or change. Inspect the draft Compare names, owners, dates, decisions, and requirements with the source. Remove or correct unsupported claims. Choose the handoff Share only a reviewed result that is appropriate for its audience. If a claim cannot be verified, return to the source owner rather than treating the AI output as evidence. Check the information rules before trying it Information and review See how to check tool approval, information class, and the right human review. Information and review Tool approval is not blanket permission Check both the tool and the information before you enter anything. An assistant being available to employees does not mean every task or information type is approved for it. Consult your company's approved-tool register and information-classification guidance. Follow the specific scope and conditions shown there; do not infer permission from a tool's presence or from another team's use. If either the tool's approval status or the information's classification is unclear, pause. Ask the role responsible for the relevant tool or information guidance, or use the company's designated AI help route. Do not paste confidential source material into a prompt to find out whether it is allowed. A practical check before sharing If a check fails or is uncertain, do not proceed on assumption. Check What to verify If unclear Tool The assistant is listed as approved for this task. Ask the tool owner or designated AI help route. Information The source is permitted under its classification guidance. Ask the information owner or follow the classification guidance. Output Names, dates, owners, decisions, and requirements match the source. Correct it from the source or ask its owner. Sharing The audience can receive the reviewed content. Ask the content owner before sharing. Common questions A tool can assist with a draft; responsibility for using and sharing it remains with people. Can I use a tool just because it is available? No. Check the approved-tool register and the information-classification guidance for the intended task and source. Availability alone does not approve every use. What if the assistant supplies an owner or decision that is missing from my source? Treat it as unverified. Remove it or check with the person or role responsible for the source. Do not present an inferred detail as an agreed fact. How can I share a useful learning? Share a short description of the task, the prompt pattern, and a lesson such as “ask the assistant to flag missing owners.” Leave out confidential source text, personal details, and unreviewed output. Use the company's designated AI help route when a question needs an authoritative answer.