I’ve built enough pitch decks at this point to realize that the hardest part is usually not making the slides look good. That part is actually pretty easy now. The difficult part is deciding what exactly should be on every slide, what should NOT be there, what order actually makes sense, and how to explain the whole company without turning the deck into either a 40 page corporate presentation or some beautiful collection of absolutely meaningless slogans.
For actually building the deck, my personal top choices are Gamma, Canva and PowerPoint. Gamma if I want to get from nothing to something presentable extremely fast. Canva if I want more control over the visual part without spending forever on it. And PowerPoint when I want complete control and know that I’m probably gonna move every single element around 25 times anyway.
But before I even open any of them, I usually work on the structure with ChatGPT or Claude. And no, I don’t mean “create me a pitch deck for my AI startup”. That’s probably the easiest way to get the most generic pitch deck imaginable.
Here are the prompts I’d actually use.
1. First, make AI understand the company before it writes anything
I’m going to give you everything I currently know about my company. Do not create the pitch deck yet. First, interview me as if you were an experienced early stage investor who is trying to understand the company properly. Ask me only questions that would materially affect the pitch deck. Focus on the problem, customer, product, why now, market, business model, traction, competition, distribution, team and what exactly I’m raising money for. If my answer is vague, push me for something more concrete. Do not accept generic startup language.
This is probably the most important one because otherwise AI loves filling the missing parts with startup bullshit. Suddenly your company is “revolutionizing the way businesses unlock value through an AI powered ecosystem” and you don’t even know what the hell happened.
When the customer story is still a guess: Use my Customer Interview Framework before answering the AI’s questions. Enter the customer you need to speak with, what you need to learn, and the assumption you want to test. The guide helps you ask about what people actually did, so the problem slide can draw on real conversations instead of AI filling the gaps.
2. Build the actual narrative, not the slides
Based on everything I told you, create the narrative of the pitch deck before creating individual slides. Explain what the investor needs to understand first, what question naturally appears in their head after that, and what the next part of the story should answer. The deck should feel like one continuous argument rather than 12 independent slides. Point out anything in my story that is currently weak, confusing, unsupported or simply not interesting enough.
I like doing this separately because people obsess way too much over whether “Market” should be slide 5 or slide 6 while the actual story makes no sense.
When the story could describe any competitor: Fill out the Product Positioning Framework with your specific customer, what they use now, the difference that matters, and proof you can show. Use that contrast to sharpen the narrative before you decide where any slide goes.
3. Now turn it into slides
Turn this narrative into a concise investor pitch deck. For every slide give me: Slide title The ONE thing the investor should understand from this slide Exact content that should appear on the slide What should be shown visually instead of written as text What I should explain verbally and therefore should NOT put on the slide Keep every slide focused on one idea. If a slide needs five different messages, split it. If information does not help the investor understand the opportunity or make a decision, remove it.
That last part is important. LLMs have the same problem as humans here. If you give them information, they feel obligated to somehow put all of it into the deck.
When you have a slide list: Put your meeting length, slide count, opening, and time reserved for questions into the Pitch Deck Timing Planner. It shows how much speaking time is left per slide. If a slide needs two minutes and the plan gives it 30 seconds, cut it, split the conversation, or move detail into what you say aloud.
4. My favorite prompt for removing AI crap
Go through this pitch deck and find every sentence that sounds like it was written by ChatGPT, a consultant, or somebody trying to sound impressive. Especially phrases like “revolutionizing”, “seamless”, “empowering”, “next generation”, “unlocking”, “transforming”, “innovative solution” etc. Replace them with the simplest possible explanation of what we actually mean. Do not make the writing more elegant. Make it more specific.
You can actually run this prompt several times. Somehow there is always another “seamless experience” hiding somewhere.
When the slide copy is drafted: Paste the text into the AI Slop Detector to catch stock phrases and repetition your own eye may skip. Treat its flags as an editing checklist, then replace vague claims with the specific customer, action, or result you actually mean.
5. Make the deck understandable to someone who knows nothing about your industry
Pretend you are a smart investor who knows absolutely nothing about this industry. Go through the deck slide by slide and tell me exactly where you stop understanding what I’m talking about. Do not assume knowledge of our terminology, competitors, workflow or technology. For every confusing point tell me what question appears in your head.
This is particularly useful when you’ve been building the product for a year and now think some completely obscure industry concept is common knowledge.
When the deck makes sense on the page: Try the opening out loud in the Pitch Workbench. Choose “Investor”, paste the words you would actually say, and use the timer and opening feedback to see whether someone can understand the company before you reach slide three.
6. Then do the exact opposite
Now pretend you are an investor who has seen 50 companies in this exact space. Tell me what in this deck sounds obvious, generic, exaggerated or identical to what every competitor would say. What would make you think “yeah, I’ve heard this before”? Do not rewrite the deck yet. Just identify the weak points.
This one hurts more, which is probably why it’s useful.
7. The brutal investor version
Act as a skeptical investor who does not want to invest. Your goal is to find reasons to reject this company. Go through the pitch deck and list the strongest objections, missing evidence, suspicious assumptions and questions I would probably get during the meeting. Prioritize the objections that could actually kill the deal rather than minor details.
I’d absolutely do this before sending the deck anywhere important. Much better to get destroyed by Claude at 2am than by an actual investor the next morning.
8. And finally, make sure every slide earns its existence
For each slide answer one question: “What breaks if I delete this slide?” If the answer is “not much”, suggest deleting it or merging it with another one. The deck should contain the minimum number of slides necessary to make the full argument.
That’s probably the closest thing I have to a final cleanup.
As for the actual tools, Gamma is probably my first choice when I need a decent looking first version ridiculously fast. I wouldn’t necessarily trust it to make the final deck exactly how I want it, but for getting the whole thing out of your head and onto slides it’s extremely convenient.
Canva is somewhere in the middle for me. Much more visual freedom, still very fast, lots of existing layouts, and you don’t really need to know anything about design to make something that doesn’t look terrible.
And then there is PowerPoint, which is still PowerPoint. Less magical, more annoying, but if I actually care about every pixel, every chart, every transition and want zero restrictions, I’ll probably end up there sooner or later.
I would not ask any of these tools to invent the pitch for me though. Let ChatGPT or Claude help you think through the company first. Then let the presentation tool make it look good. Otherwise you can very efficiently create a beautiful deck saying absolutely nothing.