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

Hey IH! I'm Nicola, founder of LimTech and AI developer from Italy.

While buying and searching for products online, I kept noticing
the same problem over and over: ecommerce stores losing customers
simply because they couldn't find the right product. Sometimes the
search bar doesn't work properly, categories are unclear, or you
don't even know exactly what you need to solve your problem.

So I decided to fix it. Here's what I built, why, and honestly —
how hard it was.

We've all been there.

You land on an ecommerce store looking for something specific.
You type in the search bar. You get hundreds of results but maybe
none of them is what you're actually searching for. You filter,
scroll, filter again. Ten minutes later you give up and leave.

Or worse — you find something close, but you're not sure if it
fits your needs. There's no one to ask. You're forced to send an
email or call. If you're lucky there's a chat that says "we'll
reply within 24 hours." So you close the tab — because you wanted
to buy right now, not tomorrow.

This frustrated me every single time. Not just as a customer,
but as someone who builds tech for businesses. I kept thinking:
why is this still a problem in 2026?

The technology to solve it exists. But nobody had applied it
properly to small and medium ecommerce stores.

So I started testing. And after months of work, I built Nexus.

Nexus is an AI assistant that reads your entire product catalog
via API — prices, availability, variants, categories — and helps
customers find exactly what they're looking for through a natural
conversation.

Customer: "I'm looking for a running shoe under €100, waterproof, size 42"
Nexus: here are the 3 products that match, with prices and
availability. Which one interests you?

No more endless scrolling. No more lost customers.
Like talking to a real shop assistant — but available 24/7.

The honest part: it was much harder than I expected.

I want to be real about this, because I see too many "I built X
in a weekend" posts. This took me several months, and there were
moments I genuinely wasn't sure it would work.

Problem 1: The server never slept

The first big challenge was automation. Nexus needs to update
its knowledge every night automatically — reading new products,
prices, availability from the catalog. Setting up reliable
server-side automation that actually runs at 2am without failing
was way more painful than I expected. Cron jobs, error handling,
retry logic, monitoring — things that sound simple but eat weeks
when you're building solo.

Problem 2: Hallucinations nearly killed the project

This was the real nightmare. Early versions of Nexus would
confidently recommend products that didn't exist, invent prices,
or suggest sizes that were out of stock. For a generic chatbot
that's annoying. For an ecommerce assistant, it's catastrophic —
imagine a customer ordering something the AI made up.

I spent weeks on this. The solution was moving to a full RAG
(Retrieval-Augmented Generation) architecture — instead of letting
the AI generate answers from general knowledge, it can only respond
based on the actual catalog data it has retrieved. If a product
doesn't exist in your store, Nexus says so clearly and suggests
real alternatives.

Zero hallucinations. Every answer is grounded in verified data.

Problem 3: Token costs almost made it unviable

The third wall I hit was cost. Every customer conversation uses
tokens — and with a large product catalog, the early versions were
sending huge amounts of data to the LLM with every single query.
The costs were unsustainable.

I had to completely rethink how Nexus retrieves and filters
information before sending it to the model. Better vector search,
smarter context windowing, aggressive filtering before the LLM
even sees the data. It took weeks of optimization to get the cost
per conversation to a level that made business sense.

Looking back, these three problems taught me more about building
real AI products than any tutorial or course ever could.

Where we are now

• Live at limtech.ai — you can try the demo right now
• Two plans: Web (reads your website content) and Ecommerce
  (full catalog API integration)
• Compatible with Prestashop, WooCommerce, and other platforms
• Live in 24 hours from setup, no technical team required
• GDPR compliant, servers in Europe

Currently onboarding our first customers and actively looking
for feedback from ecommerce store owners.

What's next?

Honestly, this is the part where the real work begins.

The product works. The product is live. But building something
and selling it are two completely different skills — and I'm
starting the second one now.

Right now I'm focused on three things:

• Finding the first paying customers

• Growing organic search. I've spent the last weeks optimizing
  limtech.ai for SEO — building the right pages, fixing technical
  issues on the server (more painful than expected), and trying
  to rank for the right keywords. Early days but the foundation
  is there.

• Learning to sell. After months of heads-down building, I now
  have to get out there and talk to real customers. If you've
  been through this transition from builder to seller, I'd love
  to hear how you approached it.

If you have suggestions, feedback, or just want to try the live website — I'm here in the comments.

What's the hardest part of going from "product is ready"
to "first paying customer"? Would love to hear your experience.