A practical guide for executives who spend half their week inside documents.
Here is a quiet problem nobody puts on a slide.
The same tool that just saved you three hours on a proposal also made it sound like everyone else’s proposal. You read it back, and it’s competent. Polished, even.
But it doesn’t sound like you. It sounds like a confident stranger who has never met your client, never sat through that uncomfortable Q3 review, never learned the hard way what this particular buyer actually cares about.
Many capable leaders have gone through this over the last year. They start using AI for the heavy lifting; the output gets faster and smoother — and somewhere in the smoothness, the thing that made their writing theirs gets sanded off. Clients notice. Boards notice. You notice.
The good news is that this is entirely avoidable. The fix isn’t to use AI less. It’s to use it differently. After a fair amount of trial and error — mine and other people’s — here is what actually works.
First, change the job you’re hiring AI for.
Most people use AI as a ghostwriter: “Write me a proposal for X.” That’s exactly the prompt that produces forgettable, interchangeable prose.
The better mental model is AI as a sharp, fast, slightly junior collaborator — one who is brilliant at structure, summarising, and stress-testing, but who has zero knowledge of your relationships, your judgement, or your taste. Your job is to supply those three things. Its job is to handle the scaffolding so you can.
Once you make that shift, everything below follows.
Part 1: The prompting moves that actually work.
Give it the skeleton, not a blank page. AI is far better at improving an idea than at originating one. Instead of “write a proposal,” spend ninety seconds dumping your raw thinking — bullet points, half-sentences, the three things you absolutely must land — and then ask it to organise. A useful prompt:
“I’m going to talk through my thinking, unstructured. Don’t add ideas of your own. Organise what I say into a clean proposal structure, keeping my wording and phrasing wherever possible. At the end, flag anything important that seems missing.”
This one technique — brain-dump, then structure — is the single biggest difference between output that sounds like you and output that sounds like a chatbot. The ideas stay yours. The AI only tidies the room.
Always hand over context, not just a task.
A request with no context gets you generic writing because you’ve given it nothing to be specific about. Before you ask for anything, tell it: who’s reading this, what decision you want them to make, what they already believe, and what could go wrong.
Compare “write an executive summary for a logistics automation project” with “write an executive summary for our CFO, who is sceptical of capex right now, for a ₹2.4 crore logistics automation project where the real argument is the 14-month payback, not the technology.” The second produces something usable. The first produces filler.
Treat it as iterative, not one-shot. The first draft is a starting position, not a finished product. Expect to go three or four rounds. That’s not a failure of the tool — it’s how the tool is meant to be used.
Part 2: Editing with AI — the underused superpower.
Here’s where most executives leave the real value on the table. AI is excellent at editing — often more useful as a critic than as a writer. And when you write the risky first draft yourself and let AI sharpen it, your voice survives intact, because it was there from the start.
A few prompts I’d keep in your back pocket:
The tightening pass: “Don’t rewrite this. Tell me where a busy reader would lose patience, which claims need evidence, and what I can cut to make it 20% shorter without weakening the argument.”
The red team: “You’re the procurement head at the client. Read this proposal and list every objection or doubt you’d raise.” This is worth more than ten rounds of polishing — it tells you where your case is actually thin.
The reverse outline: “Summarise the argument of each paragraph in one line.” If the summary reads like nonsense, your logic has a hole. Better to find it now than in the boardroom.
Notice that in all three, you wrote the document. AI is pressure-testing it, not replacing it.
Part 3: Keeping your voice (the part everyone gets wrong)
Learn the tells. AI writing has a fingerprint, and once you can see it, you can’t unsee it. Watch for the vocabulary — delve, leverage, robust, seamless, landscape, navigate the complexities of. The phrases — “in today’s fast-paced business world,” “it’s not just X, it’s Y,” “the key takeaway.” The rhythm — everything arranged in tidy threes, perfectly symmetrical bullets, an em-dash in every other sentence.
And above all, the hollow confidence: lots of assured words, but not a single specific number, name, or date. When you see those, you’re looking at a draft that hasn’t been touched by a human who actually knows the subject.
Train it on you: Paste two or three things you’ve genuinely written — an old email, a memo, a past proposal — and instruct it: “Study my sentence length, vocabulary, and how I open and close. Now draft this in that same voice.” It’s not perfect, but it pulls the output noticeably closer to home.
Inject what AI can’t know: This is the real secret to authenticity, and it’s almost embarrassingly simple. Add the specifics only you have.
The actual figure from last quarter. The client’s name and the thing they said in the last meeting. The detail you noticed on the site visit. AI can write a smooth sentence about “improving operational efficiency.” Only you can write “the third shift at the Manesar plant is where the margin is leaking.” That sentence is unforgeable — and it’s what people remember.
Read it aloud before it goes out. If you stumble, or if a sentence sounds like a press release rather than something you’d actually say across a table, rewrite it. The mouth catches what the eye misses.
Keep one weird sentence. Leave in at least one line only you would write — a turn of phrase, a bit of dry humour, an unexpected analogy. It signals, unmistakably, that a real person was here.
A quick map for the three document types:
Proposals — Let AI structure the document and anticipate objections (the red-team prompt earns its keep here). Keep the value proposition and the relationship knowledge entirely yours; that’s the part that wins the deal.
Reports — AI is genuinely strong at first drafts of sections and at summarising data you feed it. But write the recommendations and the “so what” yourself. A report’s worth lives in the judgement, and judgement is the one thing you can’t delegate.
Presentations — Use AI for slide structure and to turn dense paragraphs into clean talking points. But the narrative arc — the order in which the argument unfolds and the line you want them repeating afterwards — has to come from you. Decks don’t persuade. Stories told over decks do.
The bottom line:
The aim was never to write faster. Speed is just the side effect. The real prize is that AI clears away the mechanical work — the structuring, the first-draft slog, the tightening — so you can spend your time on the part that was always yours: the thinking, the judgement, the specific knowledge no model will ever have.
A document is only a proxy for your thinking. By all means, automate the proxy. Just don’t outsource the thinking — that’s the whole reason they wanted to hear from you.
