AI Startup Funding: 8 Dos and Don’ts for Strategic AI Investments

During the California Gold Rush, most miners did not get rich. More than 90% of internet startups failed within 120 days. Even major companies like Yahoo, AOL, AltaVista, and Netscape ultimately failed. What is different about the AI revolution? This CES 2024 discussion panel featuring four investment professionals explores strategic AI investment beyond the hype.

The 8 Dos and Don’ts

1. Look for large market opportunities. The size of the addressable market determines the potential return.
2. Focus on founders. The team matters more than the initial idea, especially in fast-moving AI markets.
3. Seek proprietary workflows and algorithms. Generic AI applications are easily replicated.
4. Invest in AI that does good. Social impact creates sustainable competitive advantages.
5. Leverage AI for faster, better decisions. The best AI investments improve decision-making quality and speed.
6. Prioritize non-replicable IP. Intellectual property that cannot be easily copied provides a durable moat.
7. Avoid narrow processes. A solution that works but is not generalizable enough limits growth potential.
8. Monitor what large companies are doing. Big tech companies can quickly enter and dominate any AI market segment.

The Data Question

The panel discusses the controversial topic of training data sourced from public domains and scraping, highlighting the evolving legal and ethical landscape.

For more tech and innovation content, check out internet access tips when traveling to Europe and preparing for Europe travel.

This panel provides a framework for evaluating AI investments that separates genuine opportunity from gold rush hype.

AI Is Different This Time — But History Still Rhymes

At the end of the Gold Rush, it was not the miners who made money — it was the merchants. And at the end of the internet era, a few dominant forces emerged, even though at the beginning they were all small companies. The AI revolution echoes both of those stories, but there are two key differences this time around.

First, from the very beginning, a couple of big dominant forces have already come out. That did not happen at the start of the internet. Second, as soon as OpenAI released their product, the adoption speed was really, really fast. Speed of adoption and early dominance — those are the two things that make this moment distinct. Markets are not efficient whenever something new arrives, and the AI revolution will bring tons of opportunities. But understanding how to separate real opportunity from hype is the whole challenge.

What Venture Investors Are Actually Looking For

When a pitch deck lands in front of a venture firm, two things immediately matter above everything else.

The first is market size. Investors are going on a journey with you. If you succeed, how large a company could you build? For venture math to work, there needs to be a path to a multi-hundred-million-dollar, billion-dollar, or ten-billion-dollar outcome. Smaller businesses can be great businesses that make a lot of money — but they may not be venture-backable businesses. The scale of the opportunity has to be big enough.

The second is the founder. At the early stage, investors are really focused on the entrepreneur as a person. It is a long, long journey, and the quality of the people matters enormously. Beyond that, investors want to understand how you plan to attack the problem — what milestones you are setting, and how you will make incremental progress toward a really big mission.

The Proprietary Edge: What Sets AI Companies Apart

For AI companies specifically, investors are looking for proprietary angles. That means proprietary workflows, proprietary algorithms that complement horizontal large language models, or proprietary data sets — something that sets a company apart from one that simply says, “We have an integration with OpenAI and we’re going to build a social network around that.”

The worst-case scenario for an AI investor is backing a company that becomes obsolete overnight because of the next update from OpenAI or Google. That fear shapes how investors evaluate defensibility. They may not be able to prevent a large competitor from eating their lunch entirely, but they want to see that the company has built real defenses.

There is also a values dimension. It does not feel right to invest in technology that is going to automate people’s jobs away. The preference is to invest in technology that does good — that helps people do more, helps entrepreneurs build companies, lifts people up rather than displacing them. That is not every investor’s filter, but for some firms it is a real part of the investment thesis.

How You Use AI Internally Matters Too

One point that is super relevant and critical: even if a startup is not building a native AI application, investors want to see how the founders are leveraging AI internally. The thing that sets startups apart from incumbents is that they can move faster, run more experiments, and do data-driven decision-making. Generative AI can make that advantage even bigger — you can iterate through tons of marketing messages and product ideas far more quickly.

We are entering what could be the golden age of startups, primarily because these innovations allow entrepreneurs to do so much more with so much less. The technologies that have been building for decades are now accessible in ways they never were before. So the question for any early-stage AI founder is not just what you are building — it is whether you are using AI to help you figure out what the right thing to build actually is.

Founder Advice: IP, Customers, and Replicability

For founders launching AI businesses, the advice from a veteran entrepreneur on the panel comes down to three things.

Control your IP. If your business relies on the latest update or is just a wrapper around someone else’s technology, your business is at stake. Any migration of that wrapper to something else is super disruptive. Make sure you have something that is not easily replicable.

Get obsessed with customers. A lot of founders fall into the trap of coming out of a research project — from industry or academia — and believing they have the coolest thing in the market, without ever talking to a customer. They pitch to investors. They never talk to customers.

Avoid the reproducibility mistake. Just because a solution works for one customer does not mean it works for every customer. How reproducible is it? That is a common mistake — getting excited about the first customer, and then finding that the next customer is completely different.

Watching the Large Players — and the Data Ethics Frontier

Investors also need to stay close to what large companies are doing, because those companies are potential acquirers and potential competitors at the same time. They cannot do everything, and taking a generalized foundational model and augmenting it with proprietary data for a specific application takes a lot of work. There is real value in that work. But if your company ends up in the crosshairs of a major tech company, the advice is to move up-market toward enterprise, where there is more stickiness and more flexibility. Enterprise customers take longer to switch, even when a newer AI option is available.

On the data side, accountability is not just the responsibility of the companies building the largest foundational models — it belongs to everyone building AI products. The question everyone should be asking: what kind of data was used to train these models? Was it all consented? Was it all copyright-cleared? Consumers have a right to know what has been served to them, just like a nutrition label on a product tells you what ingredients are inside.

Scraping is not currently illegal because it has not been legislated — it only becomes illegal once regulation arrives [?]. But the ethics do not wait for legislation. Content creators have been highly impacted. Not every content creator is the New York Times, able to sue a company like OpenAI and Microsoft. There are many anonymous creators whose work has been ingested into these models and who have lost the leverage of their own creativity. Building AI responsibly means asking these questions even when the faster, easier path is to ignore them.