Somewhere along the way, a lot of teams start treating a concept note and a pitch deck as the same content just reformatted, write the note first, then chop it into slides. I built a one shot prompt system for generating both, for government and private sector meeting readiness, and the thing that took the longest to get right was accepting that they are not the same document at all. They persuade differently, they get consumed differently, and a prompt that treats them as interchangeable produces a mediocre version of both.
Two Documents, Two Completely Different Jobs
A concept note gets read. Someone sits with it, possibly alone, possibly before a meeting even happens, and works through an argument at their own pace. It has room for nuance, for a paragraph that builds a case across several sentences, for a reader to go back and reread a section if something was not immediately clear. Its job is to survive quiet, unsupervised scrutiny.
A pitch deck gets presented. Someone is standing in front of a room, and the room is dividing its attention between the slide and the speaker. A slide dense with paragraph level reasoning fails immediately in that setting, not because the reasoning is wrong, but because nobody in that room is going to read a paragraph while also listening to a person talk. A deck has to carry meaning at a glance, with the speaker supplying the connective reasoning out loud that the concept note had to supply in writing.
The mistake most generic AI writing tools make, and the mistake my own early attempts made, is generating deck content the same way you would generate note content, then just breaking it into shorter chunks per slide. That produces slides that are technically accurate summaries and practically useless, because the compression that makes something work as a spoken visual aid is a different kind of compression than the compression that makes a paragraph readable.
Building A System That Understands Which Job It Is Doing
The prompt system had to treat these as two output modes with a shared source of truth underneath, not two versions of the same text. The underlying facts, the project rationale, the numbers, the case being made, stay identical. What changes completely is how that material gets sequenced and how much of the reasoning gets made explicit in writing versus left for a presenter to carry verbally.
For the concept note side, the instructions push toward building a case the way a written argument needs to build, establishing context, stating the problem clearly enough that a reader unfamiliar with the background can follow it, then walking through the proposed approach with enough reasoning visible on the page that the document can stand on its own without anyone in the room to explain it further.
For the deck side, the instructions do almost the opposite. Each slide gets treated as a single idea, not a subsection of an argument. The system is instructed to strip out connective reasoning that a speaker would naturally provide out loud, and instead surface the one line or one visual anchor that the slide needs to carry on its own. Where the concept note wants completeness, the deck wants restraint, leaving space for the actual meeting to do the explaining that the note had to do in writing.
Where Government And Private Sector Diverge Further
On top of that split, government and private sector audiences want the emphasis in different places. A private sector deck can open faster on opportunity and upside, because the room is usually already somewhat bought in on the general premise and wants to know the specifics quickly. A government facing concept note more often needs to open with alignment, showing clearly how the proposal connects to existing policy priorities or mandates before it gets anywhere near the specifics of the approach, because the evaluator's first question is usually about fit and legitimacy, not about excitement.
That meant the one shot system could not just have two modes for note versus deck. It needed a second axis for audience context, so the same underlying project could get framed with policy alignment leading for a ministry audience, or opportunity and outcomes leading for a private sector audience, without the team having to manually rewrite that framing by hand every time.
The Actual Lesson
Meeting readiness is not a formatting problem, it is a comprehension problem, and comprehension changes depending on whether the material will be read in silence or spoken in a room. Once that distinction became the actual foundation of the prompt system instead of an afterthought, both outputs got noticeably stronger, because each one stopped trying to do the other one's job.
Specific concept notes, decks, and client details remain confidential given the sensitivity of this work. Happy to discuss the general structuring approach with anyone building similar meeting readiness systems through the proper channel.
Written by Mohammad Farhan Habib Faraz
Senior Prompt Engineer and Prompt Team Lead at PowerinAI
www.powerinai.com
Top comments (0)