Tag: copilot presentation tips

Rafaela manages investor relations at a Spanish biotech and has been using Copilot for eighteen months. By her own assessment, she went through three distinct phases. Phase one (the first three months): excitement at how fast it could draft. Phase two (the next six months): disillusionment as she realised every draft needed substantial rewriting. Phase three (the last nine months): a quiet rebuild of her workflow that has compressed her presentation drafting time by roughly 60% — and produced output her CEO has not flagged once for AI-generated voice.

The shift from phase two to phase three came from eight specific workflow changes — not a different tool, not a different model, not a different prompt template. The same Copilot. The same paid Microsoft 365 licence. Different habits.

This article is the workflow she now uses, written for professionals who have already tried Copilot, found the output disappointing, and are looking for the practical shifts that move it from “useful sometimes” to “an actual time-saver every week.” It assumes you have access to Copilot in Microsoft 365 or Copilot for the web. It does not assume you are technical.

Why most “Copilot tips” articles do not change your output

Most published Copilot tips fall into one of two categories. The first is feature lists — “did you know Copilot can do X?” — which tell you what the tool can do but not how to use it well. The second is generic advice — “be specific in your prompts” — which is true but does not give you anything to actually do differently tomorrow morning.

The tips below are workflow-level. Each one is something you can change about how you use Copilot on Monday that will measurably improve the output by Friday. They are sequenced from highest leverage (configure once, benefit always) to most surgical (verify before sending). Read them once. Apply two or three on your next deck.

Tip 1 — configure custom instructions once, benefit forever

Custom instructions are the standing notes Copilot reads silently before responding to every prompt. They sit in the personalisation panel of Copilot for the web and (in most current configurations) propagate to the Copilot panel inside PowerPoint and other Microsoft 365 surfaces.

The four fields that move output quality the most are role, audience, tone constraints, and forbidden phrases. Tell Copilot what you do, who you typically present to, what voice you write in, and which words you never want to see. This single configuration step is the highest-leverage Copilot setting most professionals never touch. Spend fifteen minutes on it tonight; benefit on every prompt you write afterwards.

Tip 2 — context-stack every meaningful prompt

For any prompt that matters — a real deck, a real email, a real briefing — write the prompt with five context layers in mind: audience, decision being made, what already exists, what to avoid, and the format you want back. You can write them as separate paragraphs or compressed into one paragraph; either works. What does not work is leaving any of the five blank, because Copilot will fill the gap with the safe-default language that produces what one client called “corporate mush.”

The minimum useful context-stacked prompt is around 80 words; the maximum useful one is around 400. Below the minimum, Copilot is guessing; above the maximum, you are writing the deck yourself.

Tip 3 — draft in passes, never in one mega-prompt

Asking for a full deck in one prompt collapses three different decisions — what to cover, how to assert each point, how to write the body — into a single guess. Copilot has to make all three at once with no opportunity for you to redirect when one is going wrong.

Better: four passes. Pass 1 — outline only. Pass 2 — slide headlines. Pass 3 — slide body, one slide at a time. Pass 4 — editorial cleanup. Each pass uses Copilot to amplify your judgement on one decision. You correct course at every step. The total time is shorter than iterating on one mega-draft, and the output quality is dramatically higher.

Tip 4 — ask for statement headlines, not category headlines

A category headline is “Q3 Performance.” A statement headline is “Q3 EBIT delivered £42m, ahead of guidance by £4m on lower raw-material costs.” The first tells the reader what the slide is about; the second tells them what the slide says. Senior readers prefer the second by a wide margin — they read top to bottom, scanning for the answer, and statement headlines deliver the answer first.

Copilot defaults to category headlines because they are the safer guess for an unspecified audience. Override the default explicitly: “Headlines must be complete declarative statements, not categories. Each headline should make the point of the slide. Maximum 15 words.” Add this constraint to every headline-generation prompt.

Tip 5 — never let Copilot write your numbers

Copilot does not maliciously invent numbers. It does sometimes round inconsistently, paraphrase imprecisely, or transpose digits between draft and final output. For a marketing post, the cost of an inaccurate number is low. For a board pre-read, the cost is the deck.

The discipline is simple: every number that appears in the final deck must be traceable to a source you trust — usually a spreadsheet you built or a data extract you ran. Paste the numbers into your prompts; do not ask Copilot to look them up. After the cleanup pass, run your eye down every figure on every slide and confirm it against the source. The 10 minutes this takes is the cheapest insurance you will buy on the deck.

Tip 6 — audit verbs in the cleanup pass

The single most reliable signal that Copilot wrote a slide is the verbs. Filler verbs — leverage, drive, unlock, enable, facilitate, optimise — appear three to five times per page in default output. Each one signals “AI default voice” to a reader who has been exposed to enough AI-generated content to recognise the pattern.

The verb audit is mechanical. Search the deck for the filler verb list. Replace each one with the specific verb that describes what is actually happening. “Leverage AI for productivity” becomes “use Copilot to draft proposals in 25 minutes.” “Drive growth” becomes “grow revenue 12% in H2.” “Unlock value” becomes “release £4m of working capital.” Specific verbs sound human; filler verbs sound like a chatbot.

Tip 7 — refine in the same conversation, do not start over

When the output is not quite right, almost every professional’s instinct is to start a new chat and re-prompt. This loses all the context Copilot has built up across your previous prompts — and forces you to rebuild the briefing every time.

The faster move is to refine in place. “Slide 4 reads as defensive — rewrite it as a confident assertion of why the risk is acceptable. Same content, different posture.” “Move the financial impact from slide 5 to slide 2.” “Give me three alternative versions of the headline on slide 1.” Copilot will adjust without losing the underlying context. Same conversation. Tighter output. Less re-briefing.

The exception is when the original brief was wrong — wrong audience, wrong decision, wrong format. In that case, start a new conversation with a corrected context-stacked prompt. But “the output is not quite right” is almost never that situation; it is almost always a refinement situation.

Tip 8 — verify everything before any senior reader sees it

The cleanup pass exists for a reason. Three checks in the final pass save you from the small but career-affecting moments where a senior reader spots something the AI got wrong and you missed.

Check 1 — voice consistency. Read every headline aloud, top to bottom. Do they sound like the same person wrote them? Sharpen the weakest two or three to match the strongest.

Check 2 — verb audit. Search for filler verbs and replace them with specific ones (see tip 6).

Check 3 — number verification. Every figure on every slide traced to your trusted source.

This is the discipline that separates AI-assisted work that holds up under senior reading from AI-assisted work that gets quietly downgraded. Twenty minutes. Every time.

Putting it together: a 90-minute Copilot deck workflow

For a typical professional deck of 9–10 slides, the workflow above produces a final deck in roughly 90 minutes, broken down approximately as: 5 minutes (custom instructions confirm), 5 minutes (context-stacked outline prompt), 10 minutes (review and edit outline), 10 minutes (headlines pass), 10 minutes (review and sharpen headlines), 25 minutes (body pass — slide by slide), 20 minutes (editorial cleanup including all three verification checks), 5 minutes (final read-through aloud). The total Copilot interaction is about 35 minutes; the rest is your editorial judgement applied to its output.

This is dramatically faster than building the same deck from scratch — but it is also dramatically faster than the typical pattern of asking for a full deck in one prompt and then rewriting four times. The discipline is what produces the time saving.

For more practical depth on the prompt-side fix described in tip 2 — context-stacking — see the partner article on how to write Copilot prompts that produce executive-grade output. For the broader structural conventions that hold any executive deck together, AI-assisted or not, see the board presentation template guide.

FAQ

Do these Copilot tips work for ChatGPT or Gemini as well?

Yes. The workflow-level discipline — custom instructions, context-stacking, drafting in passes, statement headlines, verb audits, in-conversation refinement — applies to any conversational AI that holds context across a session. The Executive Prompt Pack is written for both ChatGPT and Copilot for this reason.

How long does it take to set up custom instructions properly?

For a first pass, fifteen minutes. For a refined version that genuinely matches your role, audience, and voice, plan to revisit and edit the instructions over the first two or three weeks of using them. The right test is the prompt described in tip 1 of this article: ask Copilot “in one sentence, who am I and what do I write about?” If the answer surprises you, the instructions need work.

What if my organisation blocks custom instructions or restricts Copilot configuration?

Some enterprise Microsoft 365 deployments lock down personalisation. The workaround is to paste a shortened version of your instructions at the top of every important prompt as a manual prefix. It is more work each time, but the underlying logic — telling Copilot who you are, who you write for, and what voice to use — is the same. Speak to IT about whether instructions can be re-enabled for senior users.

Should I tell colleagues I use Copilot to draft, or keep it private?

Most professional environments now treat AI-assisted drafting the way they treat assistant-built drafts — assumed, not concealed, but rarely the headline. The relevant question for any deck is whether the analysis, the recommendation, and the editorial judgement are yours. If they are, the drafting tool is a means to that end. Be honest if asked; do not over-volunteer.

How do I know if my Copilot output is genuinely good or just looks finished?

Read it as if you are the senior reader, not the writer. Two specific tests: (1) Does the deck make me agree, disagree, or hesitate? Generic AI output produces “fine but unmemorable” — strong output produces a clear directional response. (2) Could a senior colleague tell this was AI-drafted? If you are not sure, run the verb audit and voice consistency check one more time.

Not ready for the full prompt pack? Start here instead: download the free Executive Presentation Checklist — a one-page reference covering the structural moves that hold any presentation together, AI-assisted or not.

Pick two of the eight tips. Apply them to your next Copilot session. Notice the difference in the output. Then add a third the week after. The compounding effect of small workflow improvements is the difference between AI as a curiosity and AI as a quiet force multiplier in your week.

About the author. Mary Beth Hazeldine is Owner & Managing Director of Winning Presentations Ltd, founded 1990. With 24 years of corporate banking experience at JPMorgan Chase, PwC, Royal Bank of Scotland, and Commerzbank, she advises executives across financial services, healthcare, technology, and government on structuring presentations for high-stakes funding rounds and approvals.