The AI deck is here. The investor-ready deck still isn’t.

For sixteen years I’ve been writing investor decks for a living, not the design, the argument. The narrative, the market logic, the proof, the assumptions behind the numbers. I’ve spent the past two years watching AI arrive in my corner of the world, and watching most people draw the wrong conclusion about what it changes.

The scale is real. Gamma, the largest of the AI deck generators, passed 70M users and $100M ARR in late 2025. Its users have created more than 400M documents. Add Canva, Claude, and Gemini and a founder can turn an interview transcript and a website into a polished 15-slide deck before lunch.

That is a real improvement, and I don’t dismiss it. AI has raised the floor: bad decks are less bad, messy drafts get organized, version one arrives in an afternoon. But fundraising is not won at version one, and the improvement has a side effect almost no one prices in.

Polish used to be a signal. Now it’s table stakes.

Five years ago a clean, well-structured deck told an investor something: this founder prepared. When every deck looks finished, looking finished carries no information. And investors are spending less time per deck, DocSend puts the average review under three minutes, down a quarter since 2021, and closer to two and a half minutes when the deck arrives cold. The bar did not drop because decks got prettier. It rose, because the differentiator left is the thinking.

Most decks do not fail on design. They fail because the thinking is unfinished.

A deck is a compressed argument. In under three minutes it must establish what you do, who has the problem, why it hurts, why now, why customers pick you over the alternative, what proof shows the business working better over time, and why this team wins. AI can format those questions and generate fluent answers to all of them. It cannot tell whether the answers are true, believable at this stage, or specific enough to survive a partner meeting. Fluency and validity are different properties, and a language model only guarantees the first.

The deeper problem is averaging

A model trained on ten thousand decks writes the ten-thousand-and-first. It reaches for the median structure, the median phrasing, the median claim: large problem, simple solution, huge market, perfect timing, early traction, big vision. Nothing in that arc is wrong. It is interchangeable, and venture returns do not come from the median. If you are an investor, you have watched this happen in your own inbound: more decks, cleaner decks, and less variance among them than at any point in your career.

Here is the same problem as I see it from inside the work. In our discovery process we over-collect on purpose: hours of management interviews, every document the company has. Then only 10% of it makes the page. The judgment is in choosing which 10%: the one customer anecdote that proves urgency, the one operating detail that explains the margin. AI inverts this. It keeps everything and compresses evenly, and the detail that made the company an outlier gets averaged in with the rest. I have watched original stories go into a deck generator and come out ordinary.

Two tells give an AI deck away

The first is the title test. Cover everything on each slide except the title, then read the titles in order. In a deck where the thinking is done, they read as an argument: each title a full takeaway, and the sequence alone tells the story. In an AI deck they read as labels “Market Opportunity” or as sentences that commit to nothing. Investors run this test without knowing they are running it. It is what skimming a deck in two and a half minutes actually is.

The second is slides written to be read instead of presented. A fundraising deck exists to support a live, spoken pitch: sparse slides carrying the takeaway and a few points of proof, with the rest delivered out loud. AI does not know the verbal pitch exists, so it packs the talk track onto the slide. When I see six dense bullets, each a complete thought no founder would say in a meeting, I know nobody made the call about what belongs on the page versus in the room.

The pattern repeats slide by slide

The AI market slide proves the market exists; a strong market slide proves this company has a believable path through it, which requires a point of view on budgets, adoption timing, and the wedge that gets the company in. The AI traction slide describes activity; a strong one interprets evidence—a pilot with the right customer outweighs a wall of logos, and a small cohort with strong retention outweighs a big vanity metric. The AI financial slide shows outputs; a strong one shows how the business works. In every case the missing ingredient is the same: a decision about what the facts mean, made by someone accountable for defending it.

Which points to the real risk. A messy deck is obviously unfinished. A polished AI deck looks done, structure, professional language, consistent design, while the core issues sit unresolved underneath. The founder sends it. The investor finds the gaps in the first meeting, which is the most expensive place to find them.

Where AI belongs in the process

None of this makes me anti-AI. We use it inside Lyonshare, but we pair it with over a decade of positioning and deal-making knowledge. We know what questions to ask the founder upfront and where and how to push him for more proof. We know how to look past the surface level numbers to understand the business planning argument the numbers are telling whether knowingly or unknowingly. With discernment we formulate the argument we want one single slide to make and ask AI for a version one to consider some of its copy as a baseline to be edited. AI sharpens thinking, but used in lieu of thinking produces a story that says nothing.

The work is still the work: clarify the argument, pressure-test the assumptions, decide which evidence carries weight, make the market logic specific, connect the raise to the milestones it unlocks. AI compresses the time between having the answers and having the deck. It does nothing to compress the time it takes to have the answers.

The AI deck is here. The investor-ready deck still requires judgment.

Kelly Lyons is the founder of Lyonshare, which turns complex companies and capital stories into materials investors can understand, evaluate, and believe. Lyonshare clients include both startups and funds who have raised >$10B from the world’s leading investors.

Sources: TechCrunch; BusinessWire; DocSend