Tomas Lindqvist runs an eight-person logistics team at a regional distributor. He is a clear thinker and a slow writer, and email was eating his mornings. A routine "we are switching scheduling tools" announcement once took him 40 minutes and three rewrites, and he still was not happy with the tone. One Tuesday he timed himself: between status notes, change announcements, and feedback messages, he spent just over two hours writing email. That is more than a quarter of his working day spent staring at a blank message window. He started using AI to draft the routine ones, and within two weeks his email time was down to about 40 minutes total. The catch came in week one, when a draft confidently named the wrong go-live date and he nearly sent it. That near miss taught him the real skill: AI drafts fast, but the manager verifies, and the manager owns every word.

What This Lesson Covers

This lesson is about using AI to draft the everyday team emails (announcements, status updates, requests, and feedback) faster and more clearly, without ever giving up editorial control. AI is a draft partner, not a writer. It produces a strong first draft from your input; you provide the context, verify the facts, fix the tone, and decide every final word. Email is your voice, and your name is on it.

You will learn to treat AI as a draft partner, to write a structured prompt that gets you a usable first draft, to run a three-layer verification check before you send, and to recognize the messages you should write yourself instead. There is a full before-and-after rewrite with word counts so you can see exactly where the human adds value.

The stakes are worth naming. Managers typically spend between a fifth and a third of their working time writing or editing email, much of it on routine messages that do not deserve that share. And email is where your voice reaches people who cannot see your face. A muddled message creates confusion, dents morale, or escalates a conflict that a clearer one would have settled. A good one builds trust. AI can get you to the good one faster, but only if you verify that it says what you mean and sounds like you.

AI as a Draft Partner, Not a Writer

The mental shift that made AI useful for Tomas was small but important. AI generates clear, grammatically correct prose in seconds, but it cannot read your mind. It works from your prompt and its training, which means the output is a starting point, not a finished email. Your job has four parts: give clear context and direction in the prompt, review the draft against what you actually meant, edit for tone and voice, and make the final call on every word. Skip any of those and AI stops being a partner and becomes a liability.

AI does not write your emails. It gives you a first draft so you can spend your time on accuracy, tone, and judgment instead of on the blank screen.

Structured Prompting for Email

The single biggest lever on draft quality is the prompt. A vague prompt ("draft an email about the new tool") yields vague output. A specific prompt gets you to roughly 80 percent usable on the first try. Tomas includes seven things in every email prompt:

• What the email announces, requests, or explains.
• Why (the business context behind it).
• Who the audience is (team, executives, external partners).
• Tone (warm, formal, urgent, reassuring).
• Must-include phrases, dates, or approvals.
• Length (brief, medium, detailed, or a word target).
• Call to action (what the reader should do, if anything).

Investing two or three minutes in a prompt like that saves far more time than it costs, because the draft comes back close enough to edit rather than rebuild.

The Three-Layer Verification Check

The near miss in week one taught Tomas to never send an AI draft without three quick checks, in order.

Accuracy. Are the facts correct? Names, dates, numbers, decisions, and status. AI will state a wrong go-live date with total confidence. This layer is non-negotiable: verify every fact against your own source.

Alignment. Does the draft say what you actually meant? AI can produce a clear, well-written email that subtly misses your real point. Read it and ask whether it accomplishes your intent (inform, request, encourage, hold accountable).

Tone. Is the voice right for this audience and moment? A draft can be accurate and on-point yet too breezy for an announcement or too stiff for a quick update. Adjust until it sounds like you talking to these people.

Where AI Email Works Best, and Where to Stop

AI earns its keep on routine, clear-context communication: routine announcements ("we move to the new scheduler Monday"), status updates, feedback templates, straightforward requests, and acknowledgments. These are messages where the context is plain and the stakes are low.

Pull back, and write it yourself or hand-edit heavily, when the message carries emotional or legal weight: sensitive performance feedback, conflict or correction where a misread is costly, high-stakes changes such as layoffs, crisis communication where authenticity matters more than speed, and genuinely novel situations where AI cannot know the nuance. For these, AI can at most help you structure your thinking; the words must be yours.

Everyday Use Cases on Tomas's Team

Three recurring emails show how the workflow plays out in practice, and how much time it gives back.

The weekly status note. Tomas used to spend about 15 minutes assembling the team update. Now he jots three bullets (shipped the dock-scheduling fix, blocked on the carrier API from another team, next week focused on the inventory sync) and asks AI for a warm, 150-word update built from them. The draft lands in seconds, he checks that the progress is captured and adds one team member's name to credit them, and sends in about three minutes. The structure is consistent week to week, which his team has told him makes the updates easier to skim.

The process-change announcement. When a workflow changes, the risk is rambling and over-explaining until the message gets lost. Tomas lists the change, the reason, the impact (training time, cutover date, support available), and the one concern people will have, then asks AI to structure it empathetically in about 200 words. AI is good at ordering that logic cleanly. A 25-minute writing task becomes a 5-minute one, and the message is clearer because the reasoning is laid out in sequence.

The feedback email, handled differently. When a coordinator turned in a late and shallow analysis, Tomas did not let AI write the message. He used it only to structure his thinking: he gave it the facts (three days late, weak analysis, but strong data visualization and logical structure) and his intent (specific, constructive, confident in the person), and AI returned a rough outline of what to cover. Then he wrote the actual email himself, in his own voice, adding the personal context and the relationship history AI could not know. For anything performance-related, that is the right split: AI organizes the points, the manager writes the words.

The departure announcement. When a dispatcher moved to another division, Tomas needed a note that acknowledged her contribution, kept morale steady, and told people what happened to her open work. He gave AI the facts, her main achievements, her last day, who picks up which routes, and asked for a warm but not sappy tone that kept the focus forward. The draft came back close to right: specific about her achievements, clear on the handover, appropriately positive about her move. He changed one corporate-sounding opening line and added a sentence of his own about what she had been like to work with. For departures, AI does most of the work well, and the manager supplies the personal touch and the final voice adjustment.

Worked Example: A Before-and-After Rewrite

Here is exactly how Tomas drafted his scheduling-tool announcement, the one that used to take 40 minutes. He started with a structured prompt.

Prompt: Draft a team email announcing we are switching from our old scheduling tool to RoutePlan, going live Monday June 29. Audience: my 8-person logistics team. Why: RoutePlan cuts manual entry and links to our inventory system. Impact: a 1-hour training session Thursday; the old tool turns off Friday; I am available all week for questions. Address the worry about losing saved routes (they will be migrated for them). Tone: clear and reassuring, not corporate. Keep it to about 130 words. End with a clear next step.

AI first draft (139 words):

Subject: We are moving to RoutePlan next week

Hi team,

I am excited to announce that we are switching to a brand-new scheduling platform, RoutePlan, as part of our ongoing commitment to operational excellence! Starting Monday, June 29, RoutePlan will replace our current tool and revolutionize how we schedule routes by reducing manual entry and integrating with inventory.

There will be a training session on Thursday to get everyone up to speed. The old tool will be switched off on Friday. Please do not worry about your saved routes, as everything will be handled.

I am confident this transition will be a great success and I appreciate your flexibility as we embark on this exciting journey together.

Best,
Tomas

Running the three-layer check, Tomas found: accuracy was fine on dates, but the draft said routes "will be handled" rather than the specific, reassuring truth (migrated automatically); alignment was off because the training time was missing and the migration promise was vague where it most needed to be concrete; and tone was wrong, with "revolutionize," "exciting journey," and "operational excellence" reading as corporate fluff his team would roll their eyes at. He rewrote it.

Final email (118 words):

Subject: Switching to RoutePlan next Monday, June 29

Hi team,

Starting Monday, June 29, we are moving from our current scheduler to RoutePlan. It cuts the manual data entry we all dislike and links directly to our inventory system, so fewer double-entries and fewer errors.

What this means for you: a one-hour training session this Thursday at 2pm, and the old tool switches off Friday. Your saved routes will be migrated over automatically, so nothing is lost on your end.

I will be around all week for questions, so grab me anytime. Next step: hold Thursday 2pm on your calendar.

Thanks,
Tomas

The numbers tell the story. The AI draft came back in about 30 seconds versus the 40 minutes the old way took. It was roughly 80 percent usable, and the edit took four minutes. But the four minutes were the important part: they fixed a vague promise into a concrete one, restored the missing training time, and stripped the corporate tone so the email sounded like Tomas. The final email is also shorter (118 words against 139), because the human edit cut filler the AI added to sound enthusiastic.

Catching AI Fabrication

The most dangerous failure is fabrication: AI inventing a date, a vendor name, a metric, or a target to sound authoritative. The draft reads clean and confident, so it is easy to trust. Tomas once caught a draft that referred to "our Q2 goal of 500 deliveries" when the real target was 300. The fix is a habit: for any email containing a fact, figure, date, or name, verify it against your source before sending. Read the message aloud, and ask the one question that catches almost everything: "Would I be comfortable putting my name on this exactly as written?"

The classic version of this trap is the invented proper noun. Ask AI to announce a new support partner without naming the vendor and it will cheerfully supply one, complete with a plausible-sounding company name, and the sentence will read so naturally that your eye slides right over it. Every named entity in a draft is a claim you are making on your own authority. Check them all.

Anti-Patterns to Avoid

Send-as-is without reading. Generating and firing off an email unread invites factual errors and an inauthentic voice into your team's inbox. Always read before sending.

Leaning on AI tone for sensitive messages. Hoping AI will "handle it diplomatically" produces feedback that sounds generic and misses the relationship history that makes a hard message land. Treat AI output as a structural sketch only, then rewrite in your voice.

Prompting without context. A thin prompt gets a generic draft aimed at the wrong angle. Ask for "an email about Q2 planning" without saying what you are announcing or what you need from people, and you will get a vague overview instead of a clear request. Spend the two minutes to specify what, why, who, tone, length, and constraints.

Using AI for crisis or legal messages. For layoffs, safety incidents, or anything with legal weight, AI can miss the required gravity or compliance language. A casually worded note about a workplace incident does more damage than no note at all. Draft these yourself, with HR or Legal present.

A Pre-Send Checklist

The three-layer check catches most problems. Before anything consequential goes out, Tomas runs a slightly longer version, six questions with a clear action attached to each.

• Accuracy. Is every factual claim correct: names, dates, numbers, decisions, status? If no, fix it. If you are unsure, verify against the source rather than hoping.
• Authenticity. Does this sound like you? If it does not, rewrite it in your own words. If it mostly does, minor edits are enough.
• Intent. Does the email accomplish what you set out to do, whether that is to inform, request, clarify, encourage, or hold someone accountable? If not, revise before you worry about tone.
• Tone. Is it right for this audience and moment? Too formal for a routine update, too casual for a real announcement, or not empathetic enough for difficult news all call for an edit.
• Completeness. Does the reader have everything they need to understand and act? Add missing context, make next steps explicit, and answer the obvious question you can already hear someone asking.
• Sensitivity. Could this be misread, cause offense, or escalate a conflict? If it might, have a trusted peer read it first. If it is genuinely sensitive, write it yourself.

Responsible Use and Voice

A few principles keep AI-assisted email honest. Be transparent where it helps: a simple "I use AI to help draft communications quickly, but I review and edit every one" builds trust rather than eroding it, and the same openness with customers or vendors, where it is appropriate to mention, tends to strengthen rather than weaken the relationship. Protect your authentic voice, because your team should read an email and feel it came from you, not a chatbot; if edited output still sounds generic, rewrite it. Own factual accuracy completely, since you are responsible for every claim even when AI drafted it, and fact-checking is never something to delegate. And do not use AI to draft deceptive or misleading messages, or to write anything designed to obscure information, or to fake personalization at scale by generating hundreds of rejection notes that only sound personal; the tool should make you faster, not less accountable.

One more consideration deserves its own attention. When you are writing about policy, opportunities, or decisions that affect who gets what, check the final version for inclusive language and for whether it accounts for perspectives other than your own. AI can help you find more inclusive phrasing, but it cannot tell you whether the wording reflects what you actually intend. That verification is yours.

Terms Worth Knowing

Four ideas recur through this lesson and the rest of the module. A draft partner is AI used as a collaborator that produces initial output for you to review and edit, rather than as a writer producing finished work. Factual fabrication, sometimes called hallucination, is confidently presented false information: a wrong date, an invented vendor, a metric that sounds plausible and is not real. Prompt specificity is the level of detail in your instructions, and it is the variable most directly under your control. Tone calibration is the deliberate adjustment of style and voice to fit the audience and the moment, formal for executives, conversational for peers, and genuinely empathetic for hard news.

Practice This Week

Reading about this changes nothing. Pick two of the following and do them in the next seven days.

• Draft and iterate. Take one real email you have to write this week, an announcement, a status update, or a request. Write the prompt using the seven-part structure, generate a draft, run the pre-send checklist over it, then edit and send. Afterward, note what AI did well, what you had to change, and how long the whole thing took.
• Fact-check a sent message. Pull up an email you sent recently and verify every claim in it: names, dates, numbers, decisions. Did anything slip through? What would the impact have been if it had? Then decide what your standing system is for catching factual mistakes before you hit send.
• Analyze a voice you admire. Find an email from a manager, mentor, or peer whose writing you respect. What specific words or phrases make it sound like them? What tone choices fit that audience? How would an AI version read differently, and what does that tell you about your own edits?
• Reflect on a sensitive message. Identify a recent email that needed careful handling: feedback, a correction, a difficult change. Could AI have drafted it? Where exactly would you have needed to intervene? Write down your personal criterion for when you draft yourself instead of reaching for AI.
• Build your templates. List the three to five emails you write regularly. For each, write a prompt template that captures the context, tone, and constraints, test it, and refine it until the first draft is reliably 80 percent usable. Save them. Once a template works, the next email of that type takes 30 seconds to customize instead of 15 minutes to write.

Then take two minutes on the reflection that ties it together. Think back over your past week and find one message where this lesson would have changed your approach. What would you have done differently, and what would the outcome have been? Writing that down is where the connection between concept and practice actually forms.

Related Lessons

This lesson sits inside a wider set of skills, and several others pick up directly where it leaves off.

• Preparing Meeting Agendas and Notes applies the same draft-then-verify pattern to a different artifact, and its frameworks for organizing information are the natural next step after you can structure an email.
• Writing Status Reports and Updates extends this work to longer-form communication where you are handling complex information and multiple audiences at once.
• Adapting Tone and Audience goes deeper on tone calibration, which is the layer of the verification check that takes the most practice to get right.
• Verification Workflows turns the pre-send checklist into a systematic checking process you can apply to any AI output, not just email.
• Knowing When to Override AI is the companion to the "write it yourself" boundary in this lesson: it gives you a framework for deciding when AI output should be edited, rejected, or never used at all.
• Feedback Loops and Iteration covers how to refine prompts and output over time, which is what turns a decent email template into a reliable one.

Key Takeaways

• Treat AI as a draft partner, not a writer. It produces a first draft from your input; you supply context, verify, edit for voice, and own every final word.
• Specific prompts make the difference. Include what, why, who, tone, must-include details, length, and the call to action to get a draft that is about 80 percent usable.
• Run the three-layer check every time. Verify accuracy (facts, dates, names), then alignment (does it say what you meant), then tone (right for this audience), before you send.
• Extend the check when it matters. Authenticity, completeness, and sensitivity are the three questions that catch what accuracy and tone miss on a consequential message.
• Watch hardest for fabrication. AI states wrong dates, names, and numbers with full confidence; verify every fact against your source and read the message aloud before sending.
• Use AI for routine email, write sensitive messages yourself. Announcements, updates, and requests are ideal; feedback, conflict, layoffs, and crisis messages need your own voice.
• The human edit is where the value is. In the worked example, a 30-second draft plus a 4-minute edit beat 40 minutes of blank-screen writing, and the edit is what made it accurate and authentic.
• Build prompt templates for recurring emails. Once a template reliably gets you an 80 percent draft, save it; the next email of that type takes 30 seconds to customize.
• Keep it in your voice and own the accuracy. Edited output should sound like you, transparency about AI use builds trust, and fact-checking is always yours to do.