I’ve seen thousands of pitch decks. After selling two companies and investing in over 200, including category-defining giants like Anthropic and OpenAI, I can tell you a secret from the other side of the table: the traditional pitch deck is on its last legs. It’s a static, polished performance in a world that demands raw, verifiable proof.
Investors, especially in the AI space, aren’t just listening to your story anymore. They’re interrogating your data, your process, and your defensibility. They are trained to find the holes, the weak spots, the reasons to say "no." They’re looking for the red flags that tell them you’re not one of the breakout companies they’re hunting for.
Forget the perfect narrative for a moment. Your real task is to de-risk your startup in the eyes of an investor. I’m going to give you the inside scoop on the 10 most common AI-specific red flags we look for during due diligence. Patch these holes before you even think about fundraising, and you’ll be miles ahead of the competition.
1. The "Thin Wrapper" Façade
This is the most common trap I see. A founder comes in with a slick UI built on top of a public large language model (LLM) like GPT-4 and calls it a business. They’ll talk about their "proprietary AI" when all they’ve really built is a prompt library.
We see right through this. Why? Because if your entire value proposition can be replicated by a competitor in a weekend, you don’t have a business; you have a feature. I once had a team pitch me a "revolutionary" AI content writer. When I pressed them on their tech, they admitted it was just a series of complex prompts sent to an OpenAI API. I wished them luck and passed. We’re not investing in prompt engineers. We’re investing in companies that build lasting value.
How to fix it: Your value has to come from something more than just access to a public model. It could be a unique user experience, a proprietary dataset you use for fine-tuning, or a workflow automation that is incredibly difficult to replicate. Show us a diagram of your system. Be explicit about which parts are proprietary and which are not.
2. No Proprietary Data Moat
This brings me to the next point. In AI, data isn’t just part of the business; it is the business. If you’re building your models on the same public datasets as everyone else, you’re in a race to the bottom. The biggest red flag is a founder who hasn’t thought about their data acquisition and retention strategy from day one.
Your data is your single greatest competitive moat. It’s what makes your model unique and defensible. I invested in Scale AI because I saw early on that they understood this. They weren’t just building an AI; they were building the engine to create the data that fuels all AI. That’s a data moat.
How to fix it: Show investors a clear, compelling flywheel. How does your product generate unique data as users interact with it? How does that data, in turn, improve your model and create a better user experience, which then attracts more users and generates even more data? This is the loop that creates exponential value.
3. The "Academic" Founding Team
I see this all the time: a team of brilliant PhDs from top universities who can write groundbreaking papers but have never shipped a product. They can talk for hours about model architecture but freeze when you ask about their go-to-market strategy.
Technical brilliance is a prerequisite in AI, but it’s not enough. VCs are looking for what I call "product-obsessed engineers." These are founders who are just as passionate about the user problem as they are about the technology. They understand that the best model in the world is useless if nobody wants to use it.
How to fix it: If your team is purely technical, you need to find a commercial co-founder. Yesterday. You need someone who lives and breathes customer acquisition, pricing, and sales. Your team composition should reflect a balance between technical depth and market-facing expertise.
4. Waving Hands at the "AI Magic"
"And then the AI does its magic, and we get the result." Any founder who says this to me has lost my interest. It’s a huge red flag that signals a lack of true understanding of their own technology. It tells me they’re either hiding something or they don’t have the depth to explain it.
I need to know how your system works. I don’t need a full code review, but I do need a clear, step-by-step explanation of your data pipeline, your model architecture, and your inference process. I want to know your F1 scores, your precision and recall, and how you measure performance. Vagueness is a killer.
How to fix it: Be prepared to go deep. Create a technical appendix for your memo that details your architecture. Be able to explain the trade-offs you made—why you chose one model over another, or one training method over another. This builds confidence and shows you have real command over your domain.
5. Underestimating the Cost of Scale
Many founders are seduced by the ease of getting started with a public API. What they don’t calculate is the brutal cost of inference at scale. Your per-user cost might be pennies when you have 100 users, but what happens when you have 100,000? Or a million?
I’ve seen startups with fantastic traction get crushed by their own success because their cost of goods sold (COGS) spiraled out of control. Their cloud bills grew faster than their revenue. This is a classic AI startup death trap.
How to fix it: You need to have a clear-eyed view of your unit economics from the very beginning. Model out your costs at 10x, 100x, and 1000x your current scale. Show us you have a plan to optimize these costs over time, whether it’s through model distillation, quantization, or moving to your own infrastructure.
6. Solving a Problem That Doesn't Exist
AI is a hammer, and for many founders, every problem looks like a nail. They get so excited about the technology that they build a solution in search of a problem. They create a powerful engine but have no idea who the driver is or where they want to go.
One of my portfolio companies, RemoteTeam, which was acquired by Gusto, didn’t start with AI. It started with a painful, specific problem: managing a global, remote workforce. We solved the payroll, compliance, and HR problems first. Only then did we start layering in AI to make the experience better. The problem came first, not the tech.
How to fix it: Fall in love with the problem, not the solution. Before you write a single line of code, you should have spoken to at least 50 potential customers. What are their biggest pain points? How are they solving them now? Would they pay for a better solution? Your pitch should start with the customer, not the algorithm.
7. No Plan for the Pivot
The AI space is moving at a blistering pace. The model that was state-of-the-art six months ago is now obsolete. The platform you built on might get deprecated. A new breakthrough could make your entire approach irrelevant overnight.
Founders who believe their initial plan is set in stone are naive. I’m not investing in a static plan; I’m investing in a team’s ability to adapt. I want to know what you will do when—not if—your core assumptions are proven wrong. What’s your plan B? And C?
How to fix it: Talk about the potential pivots. Show us you’ve thought about the risks. For example, "If a foundational model provider builds our core feature, our pivot is to double down on our industry-specific workflow and integration moat, which is much harder for a horizontal platform to replicate." This shows maturity and strategic thinking.
8. The "AI Talent Wars" Blind Spot
"We’ll just hire a team of top AI researchers." This is another founder fantasy. There is a brutal, global war for AI talent. Companies like Google, Meta, and OpenAI are paying astronomical salaries. You are not going to win a bidding war against them.
A red flag for me is a founder who doesn’t have a realistic plan for attracting and retaining top-tier talent. Believing you can just post a job on LinkedIn and have ML engineers flock to you is a sign of inexperience.
How to fix it: You need a talent strategy. This often means finding a unique angle. Maybe it’s your mission, the quality of the problems you’re solving, or the opportunity to have an outsized impact. Many of the best engineers would rather be a big fish in a small pond than a cog in a giant machine. Sell them on the mission and the equity, not just the salary.
9. Confusing a Feature with a Product
Many AI startups are building cool features that should be part of a larger platform. An AI-powered meeting summarizer is a feature. An end-to-end communication platform that includes summarization, action item tracking, and sentiment analysis is a product.
Investors are wary of funding features because they are easily copied or integrated by larger incumbents. We’re looking for companies that are building a platform, a system of record that becomes stickier and more valuable over time.
How to fix it: Think bigger. Show us your long-term product vision. The feature you’ve built today might be your wedge into the market, but what does the full platform look like in three years? How will you expand from your initial beachhead to build a comprehensive solution that customers can’t live without?
10. No Clear Exit Strategy
This might sound premature, but in the fast-moving world of AI, you need to be thinking about your exit from the beginning. The landscape is dominated by a few massive players who are constantly acquiring innovative startups to fill gaps in their portfolio.
Who are your most likely acquirers? What are you building that they would rather buy than build themselves? A founder who hasn’t thought about this isn’t thinking like a strategic CEO. My first company, MovieLaLa, was a social network for movie lovers. We knew from day one that our most likely exit was to a larger media or content platform. That focus helped us shape our product and eventually led to our acquisition by Gfycat.
How to fix it: Have a slide in your appendix that lists your top 5-10 potential strategic acquirers. For each one, explain why you would be a valuable acquisition for them. This shows you understand the strategic landscape and are building a company with a clear destination in mind.
So, forget the glossy 20-page deck. Instead, write a concise memo that directly addresses these red flags. Show us your work. Give us the data. Be brutally honest about the risks and your plans to mitigate them. In today’s fundraising environment, it’s not the best story that wins. It’s the most de-risked business.
Frequently Asked Questions
How can I apply this thinking to my own situation?
Start by identifying the core principle behind the opinion, not the specific example. Then ask yourself: does this principle apply to my context? If yes, test it in a small, low-risk way before going all in.
What's the most common pushback you get on this?
People often push back by citing exceptions or edge cases. And they're usually right that exceptions exist. But building a strategy around exceptions rather than patterns is a losing game for most founders.
How has this view evolved over time?
My thinking on most topics has changed significantly over the years. Early in my career, I held many conventional views that experience proved wrong. I try to update my beliefs when the evidence changes.