AI for Communication

The fastest way to use AI for communication is to stop staring at a blank message. Dump your half-formed thoughts into the chat and let AI shape them; bullet points are enough.

> Draft a short, friendly email to my colleague Priya. Notes:
> - the Q3 report is running two days late, finance data arrived late
> - new delivery date is Thursday
> - reassure her the numbers themselves are solid
> Keep it under 120 words. Sign off from "Sam".

Notice the three ingredients: a role (friendly, to a colleague), the raw facts (the bullets), and constraints (under 120 words, named sign-off). When the draft comes back you are still the author — read it once, then refine in plain English: "Make the opening less apologetic and ask if Thursday works for her review."

The same message often needs to land differently for different readers. AI is excellent at tone-shifting because the facts stay the same — only the register changes. Start with a message you have, then ask for variations:

> Here is an internal message I wrote: "Heads up team — the launch
> slipped to next week because the vendor missed their deadline.
> We're chasing them daily." Rewrite it three ways:
> 1. Formal, for the steering committee
> 2. Reassuring, for a nervous client
> 3. Casual, for our team chat channel

The intent differs even when the information does not. For an executive, emphasise the recovery plan and soften the blame; for a client, emphasise what you are doing and drop the finger-pointing; for your team, lead with the plain facts. A useful trick is to describe the feeling you want: "Make the client feel they are in safe hands" gives AI a clearer target than "make it more formal."

AI bookends a meeting: before, it turns a loose list of topics into a timed, labelled agenda; after, it turns messy notes into a summary you can act on.

> Build an agenda for a 45-minute project sync. Topics, in no order:
> contractor budget overspend; whether to push the launch date;
> quick status from each workstream; a new compliance requirement.
> For each item give a suggested time, an owner placeholder, and
> mark it FYI, Discussion, or Decision. Put decisions early.

The "Decision / Discussion / FYI" labels alone make meetings shorter, because everyone knows what each item is for. Afterwards, paste the raw notes back in:

> Turn these raw meeting notes into a clean summary:
> "marketing wants launch moved - Priya disagrees, lose momentum.
> budget 12k over on contractors, James to confirm number Fri.
> compliance - need legal sign-off before go-live, nobody owns it."
> Format as: summary paragraph, then "Decisions",
> "Action items (with owner)", and "Open questions".

A good summary surfaces ownership and open loops. Always read it against your memory — AI works only from the notes you paste, so a thin note may make it guess at an unassigned owner.

Most weak presentations are just facts in slide form. A strong one has an arc: where we are, what is at stake, and what we want the audience to do. AI is a good thinking partner for shaping that arc before you open slide software.

> Outline a 10-minute presentation to leadership.
> Goal: approval to hire two more analysts. What I know:
> team is at 130% capacity; we turned down 4 projects last quarter;
> two competitors have larger analytics teams; cost ~[BUDGET]/year.
> Give a slide-by-slide outline (title + one-line purpose per slide)
> that builds a persuasive case and ends with a clear ask.

The result is a skeleton, not finished slides. Check the story holds: does each slide earn its place and build toward the ask? If a slide is decoration, cut it. Then sharpen it: "What is the strongest objection leadership might raise, and which slide should pre-empt it?"

The biggest worry people have is that AI makes everything sound bland and corporate. The fix is to use AI as an editor of your words, not a ghostwriter — give it your draft and ask for targeted improvements.

> Here is my draft. Do NOT rewrite it wholesale. Instead:
> - fix any grammar or clunky sentences
> - flag two places that are unclear and suggest fixes
> - keep my direct, slightly informal style
> Show your changes as a short list so I can choose which to accept.
>
> [paste your draft]

Showing changes as a list keeps you in control — you accept the good suggestions and ignore the rest, like an editor's track-changes. Ghostwriting ("write this for me") gives generic output, while editing your own draft keeps you the author. To match your style on a new message, prime it with samples: "Here are two emails that sound like me: [paste]. Using that voice, draft a reply below."

AI tone suggestions are confident, but confidence is not correctness. Politeness, directness, and formality vary across cultures, and AI defaults to a generic business register unless you tell it otherwise. Give it the context it cannot guess.

> Rewrite this for a Japanese client and a German client separately.
> Japanese version: lead with relationship and context before the
> request. German version: be direct, lead with the facts. Flag
> anything you are unsure about culturally rather than guessing.

Common places AI tone advice misses the mark, and the fix for each:

The safe habit: AI drafts, you decide. For anything high-stakes — a sensitive client, a senior leader, an apology — read every word and trust your judgement.