🪴 Building meaningful AI products

If building a startup is running a marathon, building a (genuinely useful) startup in generative AI is an equivalent to finishing it under 2 hours.

Usually, your first stages look like this:

1. you make your MVP run on one of the current LLMs

2. you implement a custom RAG

3. ???

This makes you dependent on all the OpenAI/Anthropic updates throughout the year. If you are not careful, next native GPT feature can include entire functionality of your startup and drive you out of business.

So how do we pace the unpredictability with the potential value extracted out of AI? Here’s a couple of principles I try to guide my startup with 👇🏻

Free Notion and Unlimited AI

Thousands of startups use Notion as a connected workspace to create and share docs, take notes, manage projects, and organize knowledge—all in one place. We’re offering 3 months of new Plus plans + unlimited AI (worth up to $3,000)!To redeem the Notion for Startups offer:

1. Submit an application using our custom link and select Beehiiv on the partner list.

2. Include our partner key, STARTUP4110P67801.

Apply Here

🔎 Start with solving real problems

“How can we use AI?” is a wrong question to focus on.

“What’s a meaningful problem we can solve?” is a better one.

The best AI products don’t try to introduce entirely new habits; they enhance existing workflows. 

For example, instead of building a generic chatbot, focus on a specific task it can do better than anything else, like summarising client calls into actionable steps for sales reps.

AI succeeds when it becomes invisible, blending seamlessly into processes people already use. 

Make user’s life easier, don’t add unnecessary complexity. To find these opportunities, listen closely to users, and dig into their most frustrating pain points.

⏳ Build for tomorrow, not today

Our brains find it hard to understand exponential.

Whenever I hear the word “exponential”, I think of compounding interest over the period of 30 years:

Now imagine what would happen to the graph if we prolonged the X axis to span over 100 years. That’s what happens to AI every single year.

One of our (founders’) key responsibilities is to anticipate all the changes that will happen and focus on the important problems. Problems that will not be solved by others in the next 6 months.

📩 Get a free pitch deck

Raising soon? I’ll send you over Passionfroot’s pitch deck that raised over $3M, just refer one friend via your personal referral link: https://foundersfuel.beehiiv.com/subscribe?ref=PLACEHOLDER

🌯 The secret ingredient is ….. the wrapper

What’s the deal with Apple first-time customers becoming lifetime ones?

Apple has maintained a customer retention rate exceeding 90% for the past three years, with 92% of iPhone users likely to repurchase another Apple device.

They achieve this impressive stat by focusing on the experience, rather than the product. Don’t get me wrong — their product is insane. But it’s the experience that matters to the user:

• Apple stores design,

• making unboxing memorable,

• exporting data from old device in 2 clicks,

• seamless integration of all other Apple devices,

.. this and much more creates a unique experience you never get at Microsoft.

Thinking about it now, have any of you ever been to a Microsoft store? Is there even one????

This is what makes Apple unique. What is the experience factor that makes your startup unique?

Remember: users remember how a product made them feel, not just what it did.

By sweating the details of the overall experience, you can turn a functional tool into something people love.

💧 Think deep, not wide

I wrote an entire edition on why vertical AI agents will become the future. To quickly reiterate and not to bore you:

Horizontal AI are the big LLMs that we can use for anything from asking for a cooking recipe to coding a new feature for us. It is versatile but lacks depth.

Vertical AI deals with one specific problem really well. Their dataset is all tailored to solve one specific problem, which makes it much more capable of helping you out. Going narrow helps your AI’s performance, making it more reliable and less prone to error.

Horizontal AI: ChatGPT, Claude, Perplexity
Vertical AI: Ava by Artisan, Notion AI

Depth, not breadth, drives long-term value.

✈️ How I Can Help