In the last 24 months since OpenAI broke the water of the AI space, we've gone through a handful of phases as an investor and startup community.

The first was ‘build cool models’ and a bunch of people threw money at that. The second one was ‘build cool tools to help you build cool models’. A bunch of money was thrown at that. Now, it's ‘build differentiated applications using unique data with those cool tools on top of those new models.’

As a result, the high water mark for what is not only a quality startup, but a quality product idea, has changed dramatically – and investor appetite for anything AI-based has only grown. Here’s everything you need to know to build a strong business that will resonate with VCs.

Successfully navigating the application layer

AI startups are getting acquired and put out of business everyday. So it’s valid to fear that Big Tech might see the value of your AI use case and eat your lunch. After all, what’s stopping OpenAI from folding your functionality into their model and rendering you unnecessary?

Startup founders need to be asking themselves the question that investors always ask: How do I create enough distance between myself and those incumbents that I can be sure whatever product I build is enduring rather than just ephemeral? Some food for thought:

• Don’t be a one-trick pony. You can’t just make a small modification to a model and deliver it to your customers. It’s critical to embed yourself in a workflow and solve a genuinely hard problem to win.

• Don't develop a solution and look for a problem. Find the problem and then determine whether or not AI can solve it. Doing AI for its own sake is unlikely to be a successful way to build a big business.

Optimize your long-term play

AI is at the top of every investor's agenda, but now what we are looking for is durable differentiation.

In terms of the technology, we already have a 6-12 month view of what might be changing at the architectural layer of a model. We take that into account when we speak to founders, because if your entire product roadmap is anchored to one AI model, that creates significant risk.

With respect to the product layer, we look at your early-stage mindset. Is it aligned with becoming a painkiller or a vitamin? Are you inserting yourself in a vital workflow or just creating a cool gadget? Both can make money, but only one is venture backable.

Lastly, we're looking for folks who understand that they need to meet customers where they are but not build solely for where they are. The trick is to build for 1x, but architect for 100x – so that as customer needs change and technology evolves, you don't have to constantly rewrite your system or software.

The recipe for AI startup success

There are 3 ingredients that signal huge adoption potential for investors:

1. High tolerance for imperfection. We don't have autonomous cars yet because lives are on the line and AI will need to be 100.000% accurate. But if you want a 50% head start on a contract, AI can give you that today. The less perfection required to get you in front of customers, the greener the flag.
   
   
2. Presence of ‘excess drudgery’. Think of all the lawyers who thought they’d be out setting precedents, who are actually stuck doing paperwork. If you can help them do that paperwork with AI, you’d be a welcome inclusion in an investment portfolio.
   
   
3. A structural invitation for AI adoption. Sticking with the law example, think of how many firms have to turn cases down because they don’t have the human headcount to scale. AI can help them to take on more business.

If those three boxes are checked, things get interesting. In a world where everyone's scared of AI, startups satisfying these dimensions present very compelling reasons to embrace it.

Be ready to pioneer

Last year was the year of window-shopping for AI. This is the year of slow and tepid deployments. 2025 is where we’ll start to see really meaningful stuff happen.

In the meantime, the field (and consumer tastes and preferences) will evolve so quickly that we’ll see what works vs. what doesn't much more quickly than in non-AI businesses.

My parting wisdom is twofold:

1. Prioritize customer delight above all else - this product directive remains unchanged from the “pre-AI” world.
2. Build something defensible by embedding yourself as deeply within a specific problem space as you can.

And keep as current as possible with a Section membership that includes unlimited courses and an unmatched professional community.