Soo-Jin Park manages a six-person operations team at a regional insurance company. She's a thoughtful communicator in person - her team consistently rates her 1:1s as the most useful conversations they have at work. But written communication has always taken her disproportionate time. A team-wide email about a process change that needed to land carefully could take her 45 minutes to draft. A status update to her director that needed to convey both good news and a difficult constraint took three drafts. Last year she started using an AI tool to help with first drafts. She was skeptical it would sound like her. The first draft she got back from a clear prompt didn't sound exactly like her - but it was 70% there, and she could get it to 100% in ten minutes instead of 45. That math changed how she thought about the tool.

What This Chapter Covers

Assisted Communication is the first chapter of the AI-Assisted Use track - Level 2 - and it's where most managers see their fastest, most concrete returns from AI. The four lessons cover using AI to draft team emails, structure meeting agendas and convert notes into action items, write executive-ready status reports, and adapt a single message for different audiences. Each skill is immediately applicable. You can try all four this week on real work you already have to do.

It is worth being honest about the scale of what this chapter is addressing. Studies consistently show that managers spend somewhere between 20 and 40 percent of their workweek writing: emails, reports, updates, agendas, and the endless small messages that hold a team together. That is an enormous share of your time going into work that is often routine, repetitive, and downstream of the thinking you already did. Assisted communication does not mean AI writes for you. It means AI handles the mechanical production layer, converting your thinking into a polished draft, so your attention goes to the parts that actually require judgment: the message, the audience, the timing, and the tone. Done well, it produces clearer and more consistent communication faster while keeping your voice recognizably yours.

Lesson 1 - Drafting Team Emails With AI

Team emails that land well share three qualities: the purpose is clear in the first two sentences, the tone matches the situation, and the reader knows exactly what they need to do. Most first drafts miss at least one of these. AI can help with all three - if you brief it well.

The brief is the key. A prompt like "write a team email about the schedule change" produces a generic result you'll spend more time editing than if you'd started from scratch. A prompt like "write a team email announcing that our Friday all-hands is moving to Thursday starting next week, because the conference room is booked. Tone: casual and practical. The team needs to update their calendars. Three sentences max" produces something you can use immediately.

The elements a good email brief includes: the purpose of the email, who you're writing to and what they already know, the tone you want (reassuring, direct, motivating, apologetic), what you need the reader to do after reading, and any constraints on length or format.

For sensitive emails - delivering difficult news, addressing conflict, communicating a decision the team won't like - Soo-Jin's approach is to draft the core message in her own words first, then ask AI to help with structure and word choice. Not the other way around. The human thinking needs to come first on the hard ones. AI can sharpen the language; it can't supply the judgment about what to say.

The reason Soo-Jin's math worked is that the baseline was so expensive. Drafting a clear, appropriately toned email to a team about an organizational change routinely takes 20 to 45 minutes, and a feedback email to a struggling team member takes longer. Multiply that across a week of manager email and the total is significant. For genuinely routine communication, a well-briefed AI draft cuts that time by 60 to 80 percent and often improves the result, because the structure comes out cleaner and the tone more consistent than a rushed human draft written between meetings.

That said, not all manager email is equally suited to this. Think of your email in three tiers. Routine announcements and updates sit at the easy end: the content is factual, the purpose is obvious, and the cost of a slightly misjudged tone is low, so AI can produce near-final drafts from a thin brief. Feedback and performance conversations sit in the middle and demand real personalization. AI can generate the scaffolding, but you have to verify that every behavior and example cited is accurate, that the tone matches the actual relationship, and that nothing reads as formulaic. People notice when feedback feels like it came from a template, and that noticing does lasting damage. Sensitive communications sit at the far end: restructuring news, performance improvement plans, terminations. Here AI belongs in a supporting role only, helping with structure or offering phrasing alternatives, while the core message and the final wording stay entirely yours. These messages carry legal and emotional weight that requires full managerial ownership.

The worry Soo-Jin started with, that AI drafts would not sound like her, is a real risk but a solvable one. The fix is a voice briefing you reuse: a short standing description of how you communicate, direct but warm, brief, formal but personable, whatever is actually true of you, plus a couple of your own past emails for the tool to reference. Paste it in ahead of the specific request. As you refine that briefing over a few weeks, the drafts land closer each time, and editing on routine email drops to something like two to five minutes.

The one discipline that never relaxes is review. For a routine announcement that is two or three minutes: confirm the facts, adjust a phrase, check the subject line. For anything sensitive it is a full read for tone plus verification of every specific claim. Never send an AI draft you have not read all the way through. The tool makes errors, sometimes subtle ones, and you own the email the moment you hit send.

After you get an AI draft, read it out loud. If it doesn't sound like you, that's the edit you need to make - not the content, the voice.

Lesson 2 - Preparing Meeting Agendas and Notes

Two meeting communication tasks that consume disproportionate manager time: building a useful agenda before the meeting, and converting messy notes into clean action items after. Both are good candidates for AI assistance.

For agendas, the brief needs to tell AI what the meeting is trying to accomplish, how long it is, who's in the room, and any constraints (one item is highly sensitive and needs to come last, two people need prep time before the technical discussion). With that context, AI can produce a draft agenda with time allocations and preparation notes. You review, adjust, and send. Total time: ten minutes instead of twenty-five.

Meetings are the second largest consumer of management time after communication itself, and the quality of a meeting is mostly determined before anyone walks in. Most managers build agendas too late, too vaguely, or not at all. A late agenda robs people of preparation time. A vague one produces unfocused discussion and an overrun. And the absence of an agenda is the single most reliable predictor of a wasted meeting. The most useful question to answer in your brief is the objective one: what should be true at the end of this meeting that is not true now? Everything else in the agenda follows from that answer.

Judgment matters about when to bother. AI agenda drafting earns its keep when the meeting involves multiple stakeholders with different perspectives, when the topics are complex enough that clear framing changes the discussion, when you want to send something in advance with enough structure to guide people's thinking, or when a recurring meeting has gone stale and needs a refreshed format. For a simple team check-in or an impromptu conversation, the overhead is not worth it. Use it where the quality of the agenda will meaningfully change the quality of the meeting, and skip it everywhere else. One caution on the drafts you do get back: review the time allocations against what you know about how this particular group actually talks, because the tool has no idea that your finance lead needs fifteen minutes for a five-minute item.

For notes, the input quality matters more than the prompt. Paste in your actual notes - even rough, bullet-point, shorthand notes - and ask AI to extract the decisions made, the action items with owners, and any open questions that still need resolution. The output gives you a clean summary you can share. The one essential review step: check every named action item against your own memory of the meeting. AI can misattribute ownership based on who was speaking when, not who actually took the action.

Converting raw notes is where this delivers its most dependable value, because doing it by hand takes 20 to 45 minutes of tidying that nobody enjoys. Ask for the output shape you actually need: action items with owners and due dates, decisions made, discussion highlights worth keeping, and open questions still unresolved. The structural work of organizing and formatting comes back done. What comes back unreliable is attribution and anything that depended on verbal nuance the notes did not capture.

A few habits make the conversion far more accurate, and they cost nothing once they are routine. Take notes in real time with a light structured shorthand, D for a decision, A for an action item, Q for an open question, so the tool has explicit markers to work from rather than inferring. Write attendee names into the notes so ownership can be attributed correctly instead of guessed. Review the converted notes within an hour of the meeting while your memory is still sharp enough to catch a wrong owner. And circulate within 24 hours, because the value of meeting notes decays quickly as people move on to the next thing; perfectly formatted notes that arrive on Thursday for a Monday meeting have already stopped mattering.

Soo-Jin runs her weekly team meeting with an AI-drafted agenda that she reviews and adjusts each Tuesday morning. The standing template she's built into her prompting saves her about fifteen minutes per meeting. She pastes it in, adds the specific agenda items for that week, gets a draft, and edits. The ritual keeps meetings structured without requiring her to rebuild from scratch every time.

Lesson 3 - Writing Status Reports and Updates

Status reports fail for one of two reasons: they contain too much detail and bury the key points, or they contain too little context and leave the reader unable to evaluate the information. AI can help you hit the right level - once you're clear about who's reading and what they need to know.

The input for a strong AI-assisted status report is messy and that's fine: paste in your bullet points, rough notes, metrics, team updates, blockers, and key decisions from the past week. Then give AI a clear brief about the audience. "Summarize this for my director. She needs to know: are we on track, what are the risks, and are there any decisions she needs to make? Three paragraphs max. Start with the bottom line."

The "bottom line first" instruction is important for executive-level communication. The natural impulse is to build up to the conclusion. Executives read status updates in order of urgency - if the key point is buried in paragraph three, it might not get read. A strong AI prompt that includes "lead with the most important point" will usually produce an output structured the right way.

One consistent error to watch for: AI tends to soften bad news. If a project is behind, a raw AI summary might describe it as "progressing with some timeline adjustments." That's not the same as "we're two weeks behind and need to make a decision about scope by Friday." Review AI-generated status reports specifically for whether they accurately represent the severity of any problems. Your director needs the accurate version, not the diplomatic one.

Two things are worth adding to the input pile that managers routinely leave out. The first is what is planned for the next period, because a status report that only looks backward forces the reader to ask the obvious follow-up question. The second is the sensitive context: your own note to yourself about which items can be framed positively and which need to be called out directly. None of this has to be polished. Rough bullets are fine. Your thinking is the input, and the formatted output is the part you are delegating. Once the process settles, a report that used to eat an hour or more comes down by half or better.

The same input package also lets you produce genuinely different reports for genuinely different readers without writing three times. A project team needs granular detail on tasks and blockers. A department head needs a mid-level summary organized around milestones and risks. A senior executive needs a one-paragraph snapshot with the headline metrics and the single most important risk. Specify the audience and the format in the prompt and you get all three from one assembly of raw material, which is the difference between a thirty-minute job and an afternoon.

Before any of them goes out, run four checks. Accuracy comes first: every metric, date, and attribution has to be right, because the tool does not know what actually happened, it only formats what you told it. Tone is second, and you already know the failure mode there. Completeness is third and it is the one that quietly costs managers the most: did you leave out a significant risk because including it was uncomfortable? Reports that omit important negatives destroy stakeholder trust when the issue eventually surfaces anyway, and it always surfaces. Ownership is fourth: every action item and next step needs a named owner, because a line like "we will address this" with nobody attached to it is not a commitment, it is a sentence.

For anything recurring, stop starting over. Build a template prompt once and update it each cycle with the new period's raw material. After two or three cycles you will have a workflow that takes a weekly or monthly report from an hour to something closer to 15 or 20 minutes, drafting, review, and final edits included.

Lesson 4 - Adapting Tone and Audience

The same information often needs to reach different audiences in different forms. An update that works in a team Slack channel doesn't work in a board presentation. A technical summary that's useful for your engineers confuses your operations stakeholders. Adapting content for different audiences is skilled communication work - and it's time-consuming when done from scratch.

AI handles this well. Give it a base piece of content and tell it to adapt for a specific audience and context. "Rewrite this as a 30-second verbal summary I can give my VP at the start of our check-in - she doesn't know the technical details, so use plain language and focus on business impact." Or: "Rewrite this status update for my technical lead - she needs the implementation details I left out of the exec version."

The skill here is briefing accurately on the audience. The more specifically you describe who you're writing for - their role, what they already know, what they care about, what they're going to do with the information - the more useful the adaptation. "Technical lead" is a thin brief. "My technical lead who knows our infrastructure well, is skeptical of timeline estimates, and will be deciding whether to escalate to the engineering team" is a brief that produces a much better output.

It is worth being blunt about why this matters as much as it does. The same message delivered to the wrong audience in the wrong format fails even when every fact in it is correct. An overly detailed technical update sent to a senior executive wastes their time and quietly signals poor judgment on your part. A high-level summary sent to a team that needed operational detail leaves them without enough to act on. Audience calibration is a professional competency in its own right, and what AI changes is not whether you need the skill but how expensive it is to exercise it well.

Most manager communication lands in one of four tiers, and each has a different default shape. Writing to an individual team member should be direct, specific, and action-oriented: what they own, what success looks like, where they have latitude. Concrete language, no organizational abstraction. Writing to a peer group across functions needs more context-setting, because you cannot assume anyone shares your understanding of your team's priorities or constraints, so the draft has to explain the why and not just the what. Writing to a senior leader calls for brevity, clarity, and confidence: what is happening, whether it is a concern, and what if anything you need from them, led by the headline, backed with numbers, stripped of jargon and unnecessary hedging. Writing to an external stakeholder such as a client, partner, or vendor takes a formal register, careful framing of anything that reflects on the organization, and usually some explicit relationship maintenance alongside the information itself.

The efficient way to serve all of them is a single-source workflow. Write one comprehensive master brief containing every relevant fact, decision, and piece of context, then generate from it rather than from scratch each time. Ask for an executive version that is brief and headline-led at roughly 150 to 200 words. Ask for a team-lead version at mid-detail and decision-focused, somewhere around 300 to 400 words. Ask for a full-team version with operational detail and clear actions, 500 words or more. Then review and personalize each one. The investment goes into the master brief, and every version after that costs minutes rather than a full round of writing.

Tone deserves the same explicitness as audience. The same factual message, that the project is delayed, can be delivered with urgency, calm transparency, visible concern, or spin, and those are four different messages in practice. Say what you want in the prompt: direct but not alarmist, or collaborative and we-are-in-this-together. If the first draft misses, describe what it got wrong rather than rewriting it yourself. Telling the tool "this sounds too formal, make it warmer and more direct" usually gets you there in two or three passes, which is faster than editing tone by hand.

Soo-Jin uses this for quarterly business reviews. She drafts one full-length update with all the detail, then creates two adapted versions - one for her director, one for the broader team. Three pieces of content from one source. Total additional time: about 20 minutes. What used to take her most of an afternoon takes a focused hour.

The Authenticity Check Before You Send

Whatever the format and whoever the reader, one gate sits in front of everything: does this sound like you? Read the final version and ask three questions. Does the voice match how you actually talk to people? Does it reflect your real relationship with this person or this group, rather than a generic version of that relationship? And is there anything in here you simply would not say? If any answer gives you pause, change it, even if the draft is otherwise excellent. Your name is on the message and your credibility rides on whether it reads as genuinely yours. This is the same instinct behind reading email drafts out loud, applied to every piece of communication you send.

Try This On Real Work This Week

None of these four skills need a practice sandbox, because you already have the work. Pick one real email you have been putting off and write the brief before you write anything else: purpose, reader, what they already know, tone, the action you need, and any constraint on length. Notice how much of the difficulty was actually in deciding what you wanted to say. For your next recurring meeting, draft the agenda from the objective question rather than the topic list, and afterward convert your raw notes and then check every named owner against your own memory before circulating. For your next status update, assemble the input package including next period and the sensitive context, then generate two versions for two different readers and compare what changed between them.

Then reflect on the one that felt worst. If a draft came back sounding like a stranger, your voice briefing needs work rather than your prompt. If a draft came back accurate but soft on a problem, that is the softening habit and you now know to look for it every time. If a draft came back fine and you sent it without reading it in full, that is the discipline to fix first, because it is the only one where the downside has no ceiling.

Related Lessons

Each of the four skills in this chapter has a full lesson behind it, and they are worth taking in order because the briefing habit built in the first one is what makes the rest work.

• Drafting Team Emails With AI goes deeper on the input quality problem introduced here, including how weak inputs produce generic drafts that cost more to edit than writing from scratch, and how to build the voice briefing that makes drafts sound like you.
• Preparing Meeting Agendas and Notes covers both halves of the meeting, the structured agenda that determines whether the meeting is any good and the note conversion that determines whether anything comes of it.
• Writing Status Reports and Updates works through the input package, the audience-specific formats, and the review checks that keep a fast report from becoming an inaccurate one.
• Adapting Tone and Audience takes the single-source workflow and the four communication tiers much further, including how to calibrate tone deliberately rather than by instinct.

Key Takeaways

• The brief determines the output quality. Specific prompts - purpose, audience, tone, constraints, desired action - produce immediately usable drafts. Vague prompts produce generic text that requires more editing than starting from scratch.
• Read AI email drafts out loud. If it doesn't sound like you, that's the edit to make. Voice authenticity matters most for team communications where people know you well.
• For sensitive communications, your thinking comes first. Draft the core message in your own words before involving AI. The judgment about what to say can't come from the tool. The sharpening of how to say it can.
• Verify action item ownership in meeting notes. AI can misattribute who took an action based on speaking patterns in the notes, not actual commitments. Always check named owners against your own recollection.
• Lead with the bottom line in executive updates. Build this into your prompt as an explicit instruction. Executives read status updates in order of urgency - bury the key point and it may not get read.
• AI softens bad news by default. Review AI-generated status reports specifically for whether problems are accurately represented. "Timeline adjustments" is not the same as "we're two weeks behind and need a decision."
• One source, multiple adaptations. Write one full version with all the detail, then use AI to create audience-specific adaptations. The efficiency compounds across every piece of communication you send to multiple audiences.