I built an AI that finds the right product for your customers

The same challenge faced by every fellow founder. Building and selling are two different skills.

Interesting concept. Choice overload is a real problem these days. If AI can actually narrow the options without feeling biased, that's a pretty valuable experience. Curious how you're measuring whether the recommendations are actually helping users decide.

I really liked your point that building and selling are two different skills. I think a lot of founders assume that once the product works, growth becomes a traffic problem, when it's often an activation problem.

As you start onboarding your first customers, I'd pay close attention to where they get stuck between "this looks interesting" and "I can't imagine running my store without this." Those conversations, onboarding friction points, and even the questions people ask before buying will probably teach you as much as analytics.

Looking forward to seeing the before-and-after conversion numbers once you've got a few stores using Nexus. Wishing you the best!

I like the server never sleeps problem - the next day runs for me and catching up or putting out agent fires is always fun lol !

The RAG approach to keeping product recommendations grounded is especially interesting. GeekyAnts has also been working around AI product engineering, and this is a good example of why the retrieval layer matters as much as the model.

Really enjoyed the transparency in this post, especially the part about hallucinations and token costs. A lot of AI product posts skip those challenges and make it sound like everything worked on the first try.

I also think your point about "building something and selling it are two different skills" resonates with many founders. That's probably the hardest transition.

One question: have you considered tracking the percentage of conversations that end in a product click or purchase? It seems like that metric could become one of your strongest selling points if you can show Nexus improves conversion compared to traditional search.

Wishing you the best with your first paying customers! Looking forward to seeing how the product evolves.

This is the right problem to be working on. The pattern I keep seeing from the visibility side: AI assistants only recommend products they can describe confidently, and confidence comes from consistency. If your product data, pricing, and positioning match across your site, your reviews, and every directory or marketplace you appear in, the model treats you as a safe answer. One inconsistency and it hedges or skips you entirely. Curious how you handle that on the data side: do you normalize the product info from merchants yourself, or rely on what they feed you?

This is a great premise I think

Nobody buys an AI product finder. They buy a conversion number, so pick five stores on one platform, install it free for 30 days, and measure one thing: revenue from sessions that used site search, before and after. I started Henson Group doing free migrations for months and those companies became my first paying clients, so I would park the SEO work until you have those five numbers, because a case study with a percentage sells this and a keyword ranking will not.

Really enjoyed the honest breakdown, especially the part about hallucinations and token optimization. I think many people underestimate how much engineering goes into making AI products reliable, not just intelligent. Good luck with getting those first paying customers!

So amazing and educative

The ecommerce store owner discovery challenge is really the funnel problem in disguise. You built for what you thought the problem was (product search), but customer acquisition through cold outreach feels like spam to them because you're broadcasting without proof.

What if the first customers come from places where store owners already hang out solving this problem? Shopify forums, theme communities, WooCommerce subreddits. You're still cold-reaching but the signal is different - they're already in the mindset of fixing catalog issues, not being sold to by random people.

The conversion gap arises from approaching buyers in broadcast mode (here's my solution) vs showing up where the buying signal already exists (here's how others solved this in your community).

Keep pushing on this - the product is solid but customer discovery shapes everything.

I enjoyed reading this because you talked about the hard parts instead of making it sound effortless.

The section about hallucinations stood out the most. I think many AI products underestimate how quickly trust disappears when the system confidently gives an incorrect answer. Moving to a RAG-based approach seems like the right trade-off for an ecommerce assistant where accuracy matters more than creativity.

One thing I'm curious about is your onboarding process with store owners. Once they see Nexus working, what has been the biggest reason they don't move forward? Is it trust, integration effort, pricing, or simply that they don't yet feel the pain strongly enough?

I think the answer to that question could end up shaping your go-to-market strategy more than any feature at this stage.

Wishing you the best with your first customers.

Hello Nicola, I could relate to the last part of your post the most.

Building the product has actually been the easier part for me. I'm only at the beginning of promoting and selling my own products, and I'm already realizing it's a completely different challenge. Writing code has clear answers. Marketing and sales feel much less predictable, and there's a lot more trial and error.

I also appreciated that you shared the problems you hit along the way instead of making it sound like everything came together effortlessly. Those are the kinds of posts I find the most useful.

Good luck with finding your first customers - I hope that part starts gaining momentum soon.

The shift from building to selling is probably where the assumptions get tested differently.

As you've started talking to ecommerce owners, are they responding to the same problem that originally made you build Nexus, or are they describing a different reason they'd actually pay for it?

The transition from builder to seller is the hardest reset. You spend months thinking the product is the problem — then it works, and you realize you have no idea how to sell. Talking to customers before the product is ready is the only thing that actually helps.