Prepare a board deck faster with AI: what actually works.

Nobody loses a weekend to a board deck because the slides were hard to format. They lose it because the story isn't settled, the numbers keep moving, and the deck is due Monday regardless. If you want to prepare a board deck faster with AI, start by being honest about where the hours actually go, because it isn't where most people point the tool first.

Where the hours actually go

Break a board deck down honestly and the time splits three ways: deciding what the story is, wrestling the evidence into a shape that supports it, and the layout that makes it look like it belongs in front of a board.

Most AI advice attacks the third one, because it demos well. It is also the smallest bucket. Picture a quarterly pack at a Hong Kong subsidiary reporting into a regional head office: the CFO knows the revenue number by Tuesday. What consumes Thursday and Friday is that nobody has agreed whether the story is a temporary headwind or a structural problem. That single unsettled sentence is the deck.

Prepare a board deck faster with AI: the sequence that works

The order matters more than the tooling.

• Settle the narrative. On paper, before you open anything. What is the one thing the board needs to conclude, and what three points get them there?
• Fix the structure. Section order, what each slide must prove, where the appendix starts.
• Then bring AI in. Draft the slide language, the section summaries, the transitions, against the structure you have already fixed.
• Do the human pass. Read every number and every claim as the most sceptical director in the room, because you are the one defending it.

Skip a step and it shows. A deck drafted before the narrative is settled reads fluently and says nothing. A deck polished before the argument is fixed wastes the polish on something about to change.

Where AI genuinely earns its place: a layout pass, so twenty slides don't look like three different people made them; first drafts of the translation-heavy sections, like the market backdrop or the operational update nobody enjoys rewriting each quarter; and consistency checks. That last one is the sleeper. Feed it last quarter's deck alongside this quarter's draft and ask what changed and what silently didn't. It catches more in ten minutes than a tired midnight read-through ever does. It is the same arithmetic behind automating the Excel and PowerPoint busywork: the hours come back from the assembly, not from the thinking. The prompts that do this work are four or five lines long, and there are more like them in the free prompt library.

What stays entirely human

Two things don't move. The argument, meaning the judgement about what this quarter means and what the board should do about it, is yours. You will be asked to defend it in the room, not the model. That boundary is the same one I set out at length in the honest answer on whether AI replaces your job.

And any number you will be challenged on gets checked against the primary source by a person, every time. A fabricated statistic can survive two internal reviews because it sounds exactly like the real thing: fluent, specific, and wrong. A board meeting is the wrong place to discover that.

The honest limit

AI won't tell you what your numbers mean. It will only help you say it faster once you already know.

AI will not save a deck that has no story. Feed it a pile of numbers with no narrative decided and it hands back a fluent, well-formatted deck that says nothing, and you have burned the same Friday finding out. It speeds up steps two and three. It cannot do step one.

The tell is always the same: a deck that reads beautifully in rehearsal and falls apart the moment a director asks the one question it was never built to answer.

The deck on your desk is session one

If there is a deck on your desk right now with a deadline attached, that is the best possible starting point for a working session. Your actual numbers, your actual board, not a generic example. It is the same starting point that works for analysts rebuilding where their week actually goes.

Bring it half-finished, narrative unresolved and all. That is the normal starting point, not something to fix before we begin.