Before You Hit Send: Put Your AI Work Through the 5C Check

I am always intrigued by AI; I wrote a book about it in 2019, “Alexa is Stealing Your Job,” and I still deliver webinars and workshops on AI. We know it isn’t perfect, but it can save us time.

AI can write the email and summarize the 37-page report you didn’t have time to read. It can turn your scattered notes into a beautifully organized document, draft an agenda, analyze survey comments, suggest a response to an unhappy client, and create a project update in about the time it takes you to refill your tea.

That is pretty impressive, and we should take advantage of what it can do for us. However, the one thing it cannot do is take responsibility for what you send. And that is a huge deal!

AI use in Canadian workplaces is growing quickly. Statistics Canada reports that 19.2% of Canadian businesses used AI in 2026, compared with just 6.1% in 2024. Among businesses already using AI, data analytics, text analytics, and virtual agents or chatbots are among the most common applications. (StatCan)

Employees are moving even faster. KPMG found that 51% of Canadian employees were using generative AI at work in 2025, with 79% of users reporting productivity improvements and more than half saying AI saved them between one and five hours each week. (KPMG)

I love those numbers. Saving time is exactly what we need.

But saving 20 minutes drafting an email isn’t much of a victory if the email contains the wrong date, exposes confidential information, or tells the executive team something that isn’t true.

That’s why I created The 5C Check, a functional framework for reviewing AI-assisted work before you send, share, or publish it.

The five Cs are Correct, Complete, Confidential, Context, and Consequences.

1. Correct: Is it factually accurate?

AI can sound incredibly confident while being incredibly wrong.

That combination is dangerous because polished writing feels trustworthy. KPMG’s 2025 research found that only 58% of employees who use AI always check its output for accuracy. That’s an improvement from 51% in 2024, but it still means a significant number of people aren’t consistently verifying what AI gives them. (KPMG Assets)

Imagine asking AI: “Write an email reminding employees about our vacation carryover policy.”

The email comes back beautifully written and hopefully even sounds like you. The problem? AI doesn’t necessarily know your vacation policy. It may fill in the blanks with something that sounds perfectly reasonable.

Before sending AI-assisted work, verify names, dates, numbers, policies, links, quotations, sources, and anything presented as fact.

If AI tells you the meeting is Tuesday, check the calendar. If it quotes a policy, open the policy. If it gives you a statistic, find the source.

AI can draft the fact. You still need to verify the fact.

2. Complete: What is missing?

Sometimes the problem isn’t that AI gave you something incorrect. The problem is that it didn’t give you enough.

Suppose you ask AI to summarize a 15-page proposal for your executive. It produces six excellent bullets. We need to check whether it included the deadline, the financial risk, the exception buried on page 12, and the fact that Legal hasn’t approved it yet.

Administrative professionals are often valuable precisely because we know what isn’t obvious.

Ask yourself: What details, exceptions, history, context, or stakeholders might be missing?

Your AI-generated meeting summary might list the decisions but miss the tension in the room. Your project update might look complete but omit the person waiting for approval. Your draft announcement may explain what is changing without explaining when it takes effect.

Complete doesn’t mean longer. It means nothing important has been left out.

3. Confidential: Should this information be in AI at all?

This is the C that should make us all stop for a moment. We routinely work with information other employees never see: salaries, performance reviews, contracts, customer information, board materials, legal correspondence, financial information, strategic plans, and executive conversations.

KPMG reported in 2025 that 14% of surveyed employees said they had entered proprietary non-financial company information into AI prompts, 13% had entered private financial data, and 11% had entered private client or customer information. (KPMG Assets)

Before putting information into any AI tool, ask: Am I allowed to put this here?

Don’t assume that because an AI tool is available at work, every type of information is safe to enter. Follow your organization’s approved tools, policies, privacy requirements, client agreements, and data-handling rules.

When in doubt, remove identifying and sensitive information or don’t enter it at all.

4. Context: Is this right for these people?

AI understands words. You understand relationships.

Imagine asking AI to write: “Tell my VP that her report is late and I need it by noon.”

AI might produce a perfectly professional message. But you know that your VP has been dealing with an emergency all morning. Or that this is the third time you’ve reminded her. Or that she hates long explanations. Or that sending another email isn’t nearly as effective as walking down the hall.

Context includes tone, timing, organizational culture, history, hierarchy, personalities, and relationships. This is where administrative professionals have an enormous advantage over AI. You know that technically correct isn’t always organizationally smart.

5. Consequences: What happens if this is wrong?

This may be the most important C.

Before sending something AI helped create, ask: Who could be affected if this is wrong?

A typo in the office potluck announcement is low risk. An incorrect date in a board package is a higher risk. An AI-generated response to an employee complaint would be much higher. A summary of legal advice, financial results, disciplinary action, or a contract? Stop. That may require human approval or subject-matter expertise.

The greater the consequence, the greater the level of human review required.

Think of it this way: AI can help do the work, but the person who checks the work owns the work.

Your 5C Check

Before you send, share, post, present, or publish anything created with AI, run through this quick checklist:

•        CORRECT: Have I verified the names, dates, numbers, policies, links, quotations, sources, and factual claims?
•        COMPLETE: Is anything important missing, including exceptions, background, deadlines, stakeholders, risks, or required next steps?
•        CONFIDENTIAL: Have I protected employee, client, financial, legal, proprietary, and other sensitive information, and followed my organization’s AI policies?
•        CONTEXT: Is the tone appropriate for the person, relationship, situation, timing, and organizational culture?
•        CONSEQUENCES: What happens if this is wrong? Does someone with appropriate authority or expertise need to review or approve it?

AI is going to become an increasingly normal part of our workday. In fact, Statistics Canada found that AI use among Canadian businesses tripled between 2024 and 2026. (StatCan) So our goal shouldn’t be to avoid AI, nor should it be to trust it blindly.

The opportunity is to become very good at knowing where AI ends, and our professional judgment begins. Let AI give you the first draft. Let it save you time. Let it organize, summarize, brainstorm, analyze, and suggest.

Then apply the 5C Check. Because AI may have helped create the work. But your name is the one attached to it.