The temptation is great: upload your business plan, type “Make me an investor pitch deck,” and after a few seconds look at ten neat slides. Headlines are spot on, colors are right, even the images seem well thought out. Five minutes, done. This is exactly where the problem begins.
What was a weekend of Powerpoint tinkering two years ago is now the push of a button. Tools like Gamma have taken the first step, Claude Design is now going much further because for the first time it is really consistently applying design systems instead of just producing generic, pretty slides. Your own colors, your own fonts, your own logo, a consistent visual rhythm across dozens of slides. That wasn’t possible six months ago.
The good news: foil design is no longer a bottleneck. The inconvenient consequence: If the packaging becomes trivial, the selection will be made elsewhere.
What investors really filter
Pitch decks have had a few generations of inflation behind them. They got fancier, then animated, then came storytelling workshops and pitch deck coaches. Investors have been through all of this and have not become any more naive. You know that a perfectly constructed slide says nothing about whether a business model is successful.
What they filter is the substance behind it: How sharp is the problem analysis? How credible is the founder-market fit? Does the “Why now” really work here and now or does it just sound like all the other decks right now? Is the market size honestly derived or is it Wikipedia numbers plus an optimism premium? Powerpoint never did this work, but Claude Design won’t do it either.
In other words: If 20 percent of pitch decks were visually convincing before, now 80 percent are. This shifts the selection even more towards other criteria. Therefore, founders should definitely not invest less time in their pitch deck just because the design is faster. You should simply make better use of the time you invest.
The preparatory work makes the difference
A few weeks ago I did exactly what is at the beginning of this text for my own pre-seed process: put in the business plan, formulated the pitch deck assignment, looked at the results. Visually neat. Incorrect content.
The deck sold exactly what was in the business plan. The problem: The plan was over a year old. In the meantime, the business model had changed, a co-founder had joined, and the actual investor pitch was no longer intended for the original holding company, but for a spin-off. The AI didn’t know any of this and therefore didn’t take it into account in the deck. She couldn’t know either because I didn’t tell her.
The analogy that has stuck with me ever since comes from the kitchen. Professional chefs call what happens before cooking mise en place: everything prepared, cut, weighed, ready to hand. Anyone who has done this will stand at the stove for fifteen minutes and have a good meal. If you don’t do it, you’ll stand at the stove for an hour and still end up with nothing more than spaghetti with tomato sauce.
It’s the same with pitch decks. The five minutes at the end is a nice time saver. But what happens beforehand determines the outcome.
Content and design belong in two tools
That’s the real point Claude Design makes: content development and slide design are two different steps, and they should happen in two different tools.
Powerpoint did both in one tool, but poorly. We built slides while we were still thinking about what should be on them. This was as productive as writing a strategy and choosing the font at the same time.
My approach for a few months
Step 1: Content development in dialogue with Claude
No foils, no design, just sparring. Simple: Which problem exactly, for whom? What is the robust “Why now” thesis? Where is the market size truly bottom-up and where is it just Wikipedia with interest? Which three areas in the business model do not stand up to the first critical investor look? This phase lasts two to four hours, minimum. The result is not a Powerpoint, but a structured content letter.
Step 2: Design implementation in Claude Design
Upload the letter, define the color palette and logo, have ten slides generated, fine-tune in two or three iterations. These are the famous five minutes. Although: often more like twenty, if you’re honest.
What I learned: The first step can hardly be shortened, but the second can. If you mix the two steps, you feel productive – but you’re not.
Skills: preserve experience-based sparring
A tip that many people overlook: With Claude (and similarly with ChatGPT with Custom GPTs) you can create so-called skills, small, carefully curated templates for recurring mental tasks.
I built a pitch deck prep skill that summarizes the best practices of the well-known frameworks from Sequoia, YC and Airbnb, conducts a structured interview process with the founding team and at the end spits out a master prompt for Claude Design. Things like that take time on the first pitch. It pays off on the second and third.
The skill asks me questions that I wouldn’t have asked myself a year ago. He asks about the investor persona before I write slides. He calls for a bottom-up SAM calculation. He doesn’t accept “Why now,” which essentially means “because AI is hot right now.”
In this sense, a skill is not a shortcut, but rather quality control with memory. What you have once laboriously worked out can be called up again for the next pitch.
Four places where the AI likes to take wrong turns
From my last pre-seed process, the four points where I had to readjust several times:
Outdated sources: The AI gratefully accesses everything you upload. If the business plan is a year old, she builds a deck that is a year old. She only knows what has changed in the team, model or market since then if you tell her.
Top-down market sizes: By default, something like Statista figures for the global AI market are used. Investors have read this a hundred times and are skimming it. Bottom-up is rarer and more laborious, but convincing. The AI must be explicitly forced there.
Smoothed language: Pitch deck language tends to be cliché: revolutionary, seamless, groundbreaking. The AI likes to produce this because its training material is full of it. A negative list of forbidden words in the prompt helps, concrete numbers help more.
The Too Smooth Team Slide: People like to apply here. The AI doesn’t know that your co-founder would have formulated the amount “150 FTE at Volkswagen” more cautiously. What is communicated publicly needs to be agreed upon before it is sent out, not just during the pitch.
Where the time saved belongs
If you can blow out a deck in an hour, you can produce a much better one in five hours. This isn’t a call to spend more time on the results, but rather to be honest with the time you save and invest it instead in the things that investors will really ask about.
For me these were the two or three foils that I already suspected would get stuck. Founder market fit. Why now? The gap between vision and next realistic quarter. These answers either work or they don’t. The deck around it is now the easy part.
Claude Design builds the slides. That takes it off your shoulders really well. But it doesn’t take away from what really matters. Anyone who understands this will win with the new tools. If you don’t understand it, you’ll have more stylishly designed pitch decks in the future that will be just as little funded as before. Therefore: Use the time you save for design – and put it into the business model.