How we've built our AI business before it was cool.

Hello Indie Hackers,

It has been over a year since we've started our company, Ethiack. This was before the whole hype around GenAI and its possibilities. In fact, our company went for a route a bit different from the usual niches of current AI startups - we went for cybersecurity.

Today, I want to share our learnings from building in AI

AI can be applied in many places. Don't copy what everyone is doing.

As I mentioned, we're working in the cybersecurity space. We got here because one of the founders, André, works as an Ethical Hacker and saw the opportunity. However, this is a good tip that any Indie Hacker can use.

Right now you see AI startups focusing on the same kind of niches, going after the same markets. However, there are MANY problems that can be solved through AI. We've seen great growth because we've solved painful problems with new technology. For context, we're basically replacing penetration tests. These are the standard used for security, but they often cost thousands of dollars and have variable quality. Since they're checklist based, we managed to take the same processes they follow and automate them with AI. Now our clients can have the same result, with consistent quality, available 24/7 and for a fraction of the cost. This also allows security researchers to focus on in-depth creative testing instead of doing boring tasks. So companies get better results, while pentesters get to tackle bigger challenges.

My first tip would be what industries are not getting the benefits of AI? How can I bring AI into it?

Train your own models if you want great moat, better results, and a bigger challenge

We use new technology. We had to research and train everything from scratch.

There was nothing out there that achieved the results we wanted. What this means in practice is that we have a big moat around our product, but we had to train everything from scratch, starting from an already existing model..

It also takes time to get better. Our AI is still learning. With every client that signs up, it gets a new set of assets to test and learn from. And as our clients deploy code, it can test and learn more.

Therefore, our product gets better over time, which is a disadvantage - we're slower to grow in the beginning - but it's an advantage in the future - the more we grow, the more value we add, therefore facilitating future growth. The more we specialize it, the better the results.

When we tried to use a default model, it produced many of the so-called LLM hallucinations. For example, suppose there’s no data about a specific vulnerability in our case. In that scenario, the model will simply invent information about a vulnerability that does not exist. Training the model with more vulnerability information improved the results significantly.

If you want to use a third-party model, you can get to market faster but you won't be specialized, and that can hurt you in the longer term. Besides, others can use the same model you're using, removing your moat (or at least a huge part of it). You're more vulnerable. I would say this is also why we're seeing so much churn in AI startups right now:

Market it different

I wanted to share a few words on this. I feel like Indie Hackers are not doing the best job in marketing their AI products.

If you go to our website, you'll notice we don't make AI a core of our messaging. And that's on purpose, after iterating a lot.

I think AI products often focus too much on the AI and not enough on what results it actually brings to the user. No one cares that you use AI, they just care about what you can do for them. In our case, they want better, faster, and more accurate vulnerability findings - and that's what we deliver. If they're interested in knowing how, we'll gladly explain that we use AI.

This has worked well, also proved by the fact that when we send cold emails we get 60%+ open rates and ~5% reply rate (which is great) without mentionining AI anywhere.

Main Takeaways

AI isn’t a fad. It generates value in the economy. Everyone knows that, even the big players.

However, people forget why they’re building - to solve problems. AI is just a means to an end. If it makes your customers' lives easier, by all means - use it! However, you won’t have a magical, winning product simply because you used AI.

Any questions you might have, ask. We’re here to help and share our learnings.

Thank you for reading,

Diogo