Tag: AI presentation editing

Rafaela had used Copilot to draft the strategy refresh deck. Twenty-eight slides, generated in eleven minutes, looking polished and structured. She sent it to her chief of staff for a sanity check the day before the board meeting. The chief of staff replied with a single sentence: “This reads like it could have come from any of our competitors.” Rafaela read the deck again with that comment in mind. The chief of staff was right. Every slide was technically correct. Every slide was anonymous. There was nothing in it that said this was their company, their numbers, their situation.

She had two choices. Present the deck as-is and trust that the board would forgive the generic feel because the underlying logic was sound. Or stay up that night doing the editorial pass that would convert the deck from a Copilot draft into something that sounded like senior thinking from inside the business. She chose the second. She also resented the third hour of editing, because the whole point of using AI had been to save time. But by midnight she had a deck that was unmistakably hers — and the board approved the strategy refresh the next morning without the kind of friction that usually attaches to AI-flavoured material.

The editorial pass she applied that night is not difficult. It is six specific moves, applied in a fixed order. Most senior presenters who use AI for deck drafting either skip the pass entirely (and present generic decks that get probed harder than they should be) or do parts of it ad hoc (and miss the moves that matter most). The pass is what turns AI-generated slides into board-approved slides. The board does not see the AI underneath. They see a presenter who knows the business cold.

Why boards reject raw AI-generated decks

Boards do not reject AI output because they detect AI specifically. They reject it because the same patterns AI produces — generic phrasing, evenly weighted bullets, no anchored evidence, no clear decision ask — are the patterns of presentations that historically came from junior staff or external consultants who did not understand the business. Boards have learned to push back hard on those patterns, regardless of who produced them. AI just makes those patterns appear more often, and faster, in decks that should be sharper.

Three signals trigger board scepticism almost immediately. The first is anonymous language. “Leveraging operational efficiencies to drive sustainable growth” could describe any company in any sector. The second is unanchored claims. A bullet that says “the market is shifting toward platform-based solutions” without a citation, an internal data point, or a named competitor reads as filler. The third is structural symmetry that is too clean. Three points per slide, three sub-bullets per point, three slides per section — the architecture itself signals that no human did the messy work of weighting what matters.

The editorial pass exists to remove all three signals. It does not require rewriting from scratch. It requires applying six moves in sequence. Each move targets one of the patterns boards reject. Done in order, the pass takes about ninety minutes for a thirty-slide deck. Done out of order, or partially, it takes longer and produces inconsistent results.

Move one: rewrite the headlines as findings

The first move targets the highest-leverage element on every slide: the headline. AI-generated decks tend to produce topic headlines — “Market Analysis”, “Competitive Landscape”, “Financial Performance” — because the prompt history that trained the underlying models contained mostly topic-style headlines from corporate templates. Topic headlines tell the audience what the slide is about. They do not tell the audience what to conclude. Board members do not read decks for topics. They read for findings.

Rewrite every headline as a complete sentence that states the conclusion of the slide. “Market Analysis” becomes “Three of our six target markets show declining willingness to pay for premium service tiers”. “Competitive Landscape” becomes “Two new entrants in the last quarter have undercut our pricing by twenty per cent without matching our service standard”. “Financial Performance” becomes “Revenue is on plan; gross margin is below plan by three points, driven by raw material cost inflation”.

The discipline is to make every headline answer the implicit question “what should I take away from this slide?” If the headline does not answer that question, the slide will not land. This single move usually accounts for more than half of the perceived improvement in a deck. Boards lean forward when headlines are findings. They glaze when headlines are topics.

Move two: anchor every claim to specific evidence

AI drafts will routinely produce claims without supporting evidence. “The market is consolidating.” “Customer expectations are evolving.” “Regulatory pressure is increasing.” None of these are wrong. All of them are unactionable without evidence. The second move is to read every bullet and ask one question: “What is the specific evidence behind this claim?” Then add the evidence to the bullet.

“The market is consolidating” becomes “Two of our top five competitors merged in Q3, reducing the active competitive set from twelve players to ten”. “Customer expectations are evolving” becomes “Our latest customer survey shows seventy per cent now expect same-day issue resolution, up from forty-five per cent two years ago”. “Regulatory pressure is increasing” becomes “The FCA’s new operational resilience framework, effective March, requires evidence of quarterly stress testing — currently we run annually”.

Boards trust specific evidence. They do not trust general claims. When you anchor every claim, the deck reads as if the presenter has done the work. When you leave claims unanchored, the deck reads as if the presenter has skimmed. AI cannot do this move for you, because the agent does not know which evidence is true for your specific company. This is editorial work that must be human. The most common reason AI-generated slides feel generic is precisely this absence of anchored evidence.

Move three: replace generic language with insider phrasing

Every organisation has its own vocabulary. The way your company refers to its customers, its competitors, its operating model, its strategic priorities — these are linguistic markers that signal “the person who wrote this works here”. AI does not have access to your internal language. It uses the generic corporate vocabulary present in its training data, which is the vocabulary of consulting reports, annual statements, and strategy textbooks.

The third move is to read every slide and replace generic phrases with the language your board actually uses. If your CEO consistently calls the market “the addressable opportunity” rather than “the TAM”, change every instance. If your operations team refers to incidents as “events” rather than “issues”, change them. If your customers are “members” or “clients” or “partners” — never “users” — change them. These edits are small. The cumulative effect is large. A deck written in your company’s language reads as insider. A deck written in generic corporate language reads as outsider, regardless of whether the author is the CEO.

Move four: cut the slides that exist to “sound complete”

AI-generated decks tend to produce more slides than the argument needs, because the underlying prompt usually asks for completeness. “Build a strategy refresh deck for the board” produces a deck that covers everything a strategy refresh deck might cover, including sections that are not relevant to your specific situation. The fourth move is to read every section and ask “would this section’s removal weaken the argument?” If the answer is no, remove the section.

Common candidates for cutting include macro-environment scene-setting that the board already lives inside; competitor profiles for competitors the board does not consider strategically relevant; appendices that exist because the AI defaulted to producing them; and “principles” or “values” slides that signal a strategy team’s thinking process rather than the board’s decision criteria. A twenty-eight-slide deck rarely needs to be twenty-eight slides. Eighteen well-edited slides almost always read sharper than the same content stretched across twenty-eight.

Cutting is harder than adding. AI tends to over-include. Senior judgement is what subtracts. The board will not miss the slides you cut. They will notice the cleaner argument that results.

Move five: install the decision sentence

The fifth move is to identify what the board needs to take away from the deck — the actual decision, recommendation, or judgement you want them to land on — and to install that sentence in three places: the closing line of the executive summary slide, the headline of the strategic recommendation slide, and the closing slide before any appendix. The same sentence, in the same words, in three places.

AI drafts almost never do this. They produce closing slides that summarise key themes (“In summary, the strategy refresh focuses on three priorities…”), which is not the same as installing a decision the board can act on. The decision sentence has a specific shape: a verb, a quantified action, a timeframe, and a qualifier. “Approve a phased twelve-month investment of £4.2m to consolidate the European platform, contingent on the operational checkpoint at month six.” That sentence can be voted on. “Focus on European platform consolidation” cannot.

Installing the decision sentence in three places is deliberate redundancy. The board reads slowly. Some members read only the executive summary. Some read only the strategic recommendation slide. Some read only the closing. Repeating the decision sentence guarantees that every reader sees it, regardless of where their attention lands. If you want to see how to structure these decision sentences across an entire deck, the AI-Enhanced Presentation Mastery course covers the decision-sentence pattern in module four with worked examples for board, investment committee, and executive committee scenarios.

Move six: read it aloud against the board’s likely reaction

The final move is the cheapest and the most consistently skipped. Read the deck aloud, slide by slide, and after each slide ask “what would each of the board members say to this?” Name them in your head. The CFO who probes assumptions. The chair who asks for the unintended consequences. The non-executive director who challenges the timing. The CEO who tests whether the recommendation is too cautious or too bold. For each likely reaction, ask: does the slide already address it, or do I need to add a line?

Some slides will need additional context. Some will need a caveat the AI omitted. Some will need an explicit “what we considered and rejected” line that pre-empts the board’s natural alternative-generation. These additions are small. They turn a deck that looks complete on paper into a deck that holds up live. The aloud-read also reveals language that looks acceptable on screen but sounds awkward when spoken — almost always a sign of phrasing the AI inserted that needs replacement.

This sixth move is what separates decks that get approved from decks that get parked for a follow-up meeting. The first five moves clean the deck up. The sixth move makes it land in the room.

FAQ

How long does the editorial pass take for a thirty-slide AI-generated deck?

Done in order, the six moves typically take seventy-five to ninety minutes for a thirty-slide deck. Done out of order or partially, the same work usually takes two to three hours and produces inconsistent results. The order matters because each move targets a specific failure pattern, and earlier moves clear ground for later ones to land more easily. The headline rewrite, in particular, exposes weaknesses in the underlying argument that the next moves can then address.

Can I use AI to do the editorial pass too?

Partially. AI can flag bullets that lack evidence and suggest replacements where the evidence exists in your source documents. AI cannot replace generic language with your company’s insider vocabulary, because it does not have access to your internal language. AI cannot decide which slides to cut, because the cutting decision rests on judgement about what the board actually cares about. The fastest workflow is human-led editorial pass with AI used to flag candidate fixes — not the other way round.

Will the board notice that AI was used?

Boards rarely care about the tooling. They care about whether the deck reads as senior thinking from inside the business. A well-edited AI-assisted deck will not draw any specific reaction beyond the normal probing the deck content invites. A poorly-edited AI-assisted deck will draw the same reaction as a poorly-prepared deck of any origin: probing questions about why the argument is generic. The disclosure question is a non-issue if the editorial pass has done its work. If you want the framework for handling direct AI-disclosure questions when they do arise, the three-step response structure handles them in under thirty seconds.

Does this editorial pass apply to other AI tools, not just Copilot?

Yes. The six moves are tool-agnostic. They target the failure patterns of generic AI output regardless of whether the underlying model is Copilot, ChatGPT, Claude, or Gemini. The patterns are the same because the training data overlaps. The pass works on any AI-generated executive deck.

The Winning Edge — Thursday newsletter

Every Thursday, The Winning Edge delivers one structural insight for executives presenting to boards, investment committees, and senior stakeholders. No general tips. No motivational framing. One specific technique, one executive scenario, one action. Subscribe to The Winning Edge →

Not ready for the full programme? Start here instead: download the free Executive Presentation Checklist — a single-page review you can run on any AI-assisted draft before the editorial pass.

Next step: take the next AI-generated deck on your calendar and run the six moves on it in order. Track the time it takes. Note which moves expose the weakest parts of the underlying argument. Those are the moves you will get faster at — and the ones that will most consistently produce approved decks.

Related reading: The Copilot Agent Mode workflow that produces editable executive drafts.