Discover more from Ben There Building That
How To Build An AI Business That Wins
A new frontier was just unlocked. How will you capitalize?
👋 Hey, I’m Ben! I write weekly about how to grow products and companies. I go deep on growth strategies, how to build products users love, and what actionable lessons can be learned from what best-in-class companies are doing and industry experts are saying.
Happy weekend everybody!
At this point, I’m fairly confident everyone reading this post will have played around with ChatGPT and experienced first hand how powerful it is. I read a great article this week by Jaryd Hermann (check out his substack - it’s excellent) about the AI business landscape and how OpenAI grows. It got me thinking about all the businesses that have recently launched on top of OpenAI’s APIs and how to build a business that wins amid this AI gold rush.
If you’ve visited Indie Hackers or Product Hunt in the past month you’ll notice a trend; there are a LOT of people talking about and building with AI. It’s all the rage. As a fun illustration of where AI is in the hype cycle, take a look at this Google trends chart comparing AI interest vs. crypto’s 2021 peak. There is undeniably a lot of energy and attention focused on AI right now.
What does the AI opportunity landscape look like?
In Jaryd’s article he outlines the four layers to the AI industry:
layer 0: The tech needed to actually run a model (cloud computing, chips, etc.).
layer 1: The AI model itself. There will likely only be a few of these because they are expensive as hell to build and maintain.
layer 2: People will build refined layers on top of the layer 1 models that focus on a particular use case (ex. law) which gives their product some differentiation vs. the standard layer 1 model. This will be where we see more verticalization and specialization of AI.
layer 3: Applications built on top of layer 1 and layer 2 models. Midjourney is an example of this.
The majority of the new products and applications we have seen launch over the past month fall into layer 3 of the above framework. This makes sense because it has the lowest barrier to entry. The downside here is that it also has the smallest moat which will make it difficult to raise VC funding and difficult to build a sustainable business (not that either of those things are ever easy).
How do I build in this space and win?
I’ll write a couple quick thoughts on each of the layers here but I’ll spend most of my time on layer 3 since that is where the majority of you would be considering building.
Layer 0: This is an area I’ll leave to the big guys. It’s capital intensive and there are a lot of multi-billion dollar companies that already play in this space. I think Microsoft has done a great job of positioning itself to win here by heavily investing in OpenAI. This not only gives them financial exposure to a major player in the layer 1 category but it also gives Microsoft several other significant benefits: 1) It ensures that OpenAI will exclusively use Microsoft’s Azure infrastructure, 2) it allows Microsoft to deploy OpenAI’s models across all their consumer and enterprise products, and 3) it allows them to be early to market but distance themselves from any potential reputation damage from AI’s flaws (bias, misinformation, etc.). This is the kind of strategy move that gets me pumped up.
Layer 1: This is another game I’m willing to leave to the big guys. Building, training, and maintaining AI models is insanely expensive (some estimates say OpenAI is spending $100k/day to run ChatGPT, let alone the billions they spent getting to this point). We will likely see some other products come to market here as a number of big tech companies have been working on various AI projects behind closed doors (Google, Facebook, Apple, etc.) that could start coming into the public light in the coming weeks and months after seeing ChatGPT’s success.
Layer 2: This is where I start to get excited. This is where all of the “specialized AI for X” ideas will live. Replace X with any number of exciting industry-specific use cases and there will be businesses built for it. A few that come to mind and I am confident will be big: specific areas of law, customer service, education, healthcare, etc. Fine tuning the base model to specific use cases means that there can be true product differentiation between products (and thus technological moats can start to exist); for example two layer 2s focused on the same legal problem will have different outputs because each company refined the models differently. There is a lot of opportunity for people to identify niches that would benefit from highly-specific AI and then refining the base model around that use case and building tools on top of it.
Layer 3: As I mentioned before, this is where the majority of the smaller AI tools we are seeing right now are being built. This is the easiest area to build in (OpenAI makes their APIs very accessible and the pricing is reasonable) but it also is the area that has the lowest barrier to entry which makes it difficult to build a sustainable business here. For example, unlike layer 2s, if two layer 3s built on GPT3 solve the same problem they will have very similar outputs because the models they are using under the hood are the same.
So how do I win as a layer 3?
Well, first we have to define what it means to win. There will be plenty of people that build undifferentiated products that are not sustainable long term but that do make them money. That is certainly a win, but for the sake of what I will outline here I am referring to winning as building something sustainable that can continue to survive despite fierce competition even after the AI hype cycle dulls.
There are four ways companies can win long term in this space.
Build a superior brand. This is easier said than done (building a brand is very hard) but if you are able to successfully build a brand people will default to your product vs. similar ones and even be willing to pay a premium for it.
Have superior distribution. Distribution is always a competitive advantage in business and the AI gold rush is no different. Having reach matters and will help your business grow. When leveraged appropriately, having strong distribution can also help you grow other sustainable moats like your brand, SEO value, etc.
Build proprietary tech around the AI. The AI itself may not be differentiated but that doesn’t mean you can’t build your product in a way that is. Things like a superior UX and additional functions and computations under the hood beyond just what you’re using AI for can make a meaningful difference in the product experience.
Establish strategic partnerships and integrations. Having exclusive partnerships with bigger players or a broader assortment of integrations than your competitors gives your product a competitive edge that other companies will have difficulty replicating.
I encourage you to strategize around these pillars if your goal is to build a business that has staying power.
The old “picks and shovels” adage applies to the AI gold rush as well. You don’t have to build a business with AI to make money off of the growth of the AI industry. Here are a couple ideas to get your creativity flowing:
Prompt engineering consulting services
Prompt engineering courses / guides
AI training and consulting services
AI tool directories
AI oriented newsletters, blogs, podcasts
Building a layer 3 product is (relatively) easy. Winning with a layer 3 product is harder. We are going to see a lot of cool stuff built over the next few months and years and I hope that some of the readers of this newsletter are a part of that fun. If you are, reach out to me; I’d love to hear about it!
Until next week,
If you enjoyed this brain expansion nugget, consider giving it a like, share, and if you haven’t, subscribe for free to get more insights on how to build strong businesses around products that users love.