---
title: "AI product advisor: the buyer describes the problem, the site names the model"
description: "A recommendation layer over your own catalogue and technical documents: the buyer says what they need in plain language and gets specific products back, each with the reason and a link to the real page."
canonical: https://aumcreate.com/services/ai-product-advisor
dateModified: 2026-09-09
---
# AI product advisor: the buyer describes the problem, the site names the model

> An AI product advisor is a page where a buyer describes their situation in ordinary words — the load, the site, the material, the constraint — and the site answers with specific products from your catalogue and the reason for each one. It reads your own catalogue and technical documents, not the open internet, and every answer links to a real product page so the buyer can check it. It is for catalogues where choosing correctly requires knowledge the buyer does not have and your salespeople repeat all day.

## Who this is for

You sell a technical range where the right model depends on the customer’s situation: industrial equipment, components, systems, anything sold with a specification sheet. Your sales team answers the same three questions every day, and the website answers none of them.

## The problem

The catalogue is complete and correct, and the buyer still cannot use it. He does not know whether he needs the 6 kW or the 12 kW, so he does the only thing left: sends an enquiry that says “please advise”, or leaves and asks a competitor whose page did answer him.

## What we do

1. Ground the advisor in your own catalogue and technical documents, so it recommends only products that exist, at specifications that are yours.
2. Make every answer cite a real product page. An answer the buyer cannot verify is worse than no answer, and inventing a model number is the one failure that costs a real order.
3. Write down the selection logic your engineers already use — the rules of thumb, the thresholds, the “never use this below that” — because that is what makes a recommendation right rather than plausible.
4. Keep it inside its subject: it answers about your products, and it says so when the question is outside them instead of guessing.
5. Hand the conversation to the enquiry with the recommendation attached, so your salesperson opens a request that already knows what was discussed.
6. Cap the cost. The advisor answers from your catalogue, so the expensive part stays bounded and predictable rather than growing with traffic.

## Proof

Two things to look at. The demo below is an industrial energy storage catalogue: describe a load in plain words and it comes back with a specific system and the reason. And this site runs one on itself — the assistant on our own homepage recommends a template from our real catalogue, with the same rule about linking to a page you can open. If ours invented a product, you would catch us immediately, which is exactly the standard your buyers will hold you to.

## FAQ

### Will it invent products?

It answers from your catalogue and cites the page it took each recommendation from, which is what makes an invented model visible instead of plausible. A general chatbot bolted onto a site will invent one, confidently, and your buyer will order against it.

### Does it replace our sales team?

No. It replaces the three explanations they repeat all day, so the conversation they do have starts further along. The judgement calls — the unusual site, the exception, the negotiation — were never the part a page could take over.

### What does it cost to run?

Less than people expect, because it answers from your own catalogue rather than reasoning from scratch, and because the volume is bounded by how many buyers actually ask. We put the caps in at the start rather than after the first surprising month.

### What if the catalogue changes?

It reads the catalogue you already maintain, so a new product appears in its answers once it exists on the site. Nothing is copied into a second place that then goes stale.