How lighting showrooms can get the most out of AI

An AI assistant loaded with your product catalog can answer a customer’s question in seconds, the way your best salesperson would. Getting there takes four decisions: which plan, where you work, how you feed it your catalog, and how you ask. This guide walks through each one, including the part nobody warns you about.

Step 1: Pick a plan

For a showroom team, start with Claude’s Team plan, not a pile of personal accounts. Team gives you one bill, an admin who controls who has access, and Projects your whole staff can share. Your catalog gets loaded once and everyone works from the same version.

It also matters for your data. On business plans, your conversations and uploaded files aren’t used to train the model by default. That’s worth having when you’re uploading dealer pricing.

One thing to know up front: Claude doesn’t generate photos or product renders. What it does well is turn your data into things customers can use, like spec comparisons, one-page summaries and printable PDFs. More on that in Step 4.

Plans and pricing change often. Check Anthropic’s current pricing page before you buy.

Step 2: Pick where you work

Claude runs in a web browser, as a desktop app, and on your phone. Your chats and Projects follow you across all three.

For bigger jobs, set up a Project per customer. A Project keeps the files, notes and instructions for that job in one place. Upload the floor plan, the fixture schedule and the relevant catalogs, and every chat inside it starts with that context. When the builder calls three weeks later, you’re not starting from scratch.

Step 3: Load your catalog

The AI only knows what you give it, so your product data is the whole game. Here’s how to do it by hand.

1. Export your products from your vendor portal, POS or website as a CSV or Excel file.
2. Keep one file per brand. Mixing manufacturers in one sheet means mismatched columns: one brand calls it “Finish,” another “Color,” a third buries it in the description. Separate files keep the AI from confusing them.
3. Upload into a Project so the whole team works from the same files.
4. Add images where you can, several per product if you have them. They help the AI describe style and finish.
5. Add the supporting material: spec sheets, manufacturer product guides, training decks. Some showrooms upload the full annual catalog from each manufacturer.
6. Turn off web search, or tell it to stay in your files. Otherwise the AI may go looking online and recommend products you don’t carry, or quote someone else’s price.

Where it gets painful

That works, until the first time something changes.

• It’s a snapshot. The moment a price changes, a SKU is discontinued or a new collection lands, your upload is out of date. And the AI doesn’t know it’s out of date.
• It’s every brand, every season. Twenty manufacturers means twenty exports to redo, clean up and re-upload, on repeat.
• Messy data in, messy answers out. Missing dimensions, blank lumen ratings and inconsistent finish names all show up as wrong or vague answers.
• Big catalogs hit limits. Very large files have to be split up, and the more you split, the more there is to keep in sync.
• Nobody owns it. In most showrooms, “update the AI files” becomes nobody’s job within a month.

Step 4: Ask like a customer would

Once your catalog is loaded, this is where it pays off. You don’t need special phrasing. Ask the AI the way a customer asks your staff:

• “Indoor pendants for a kitchen island, under $500, white finish”
• “Dimmable pendant for a 10-foot ceiling over a 6-foot island, brass, 2700K”
• “Luxury bathroom vanity lighting for a 48-inch mirror”
• “Outdoor wall lanterns rated for coastal salt air, black, under $300”

Be explicit. The more detail you give, the better the match. Ceiling height, room size, finish, color temperature, budget and style all narrow the results. The AI is good at connecting details across thousands of products, as long as those details are in your data.

Turn answers into something to hand over. Claude can build what your customer walks out with:

• A side-by-side comparison of three fixtures with specs and pricing
• A one-page summary of everything selected for a room
• A printable PDF take-home or a fixture schedule a builder can use
• A clean email recap after a consultation

What used to be an hour of copy-and-paste after the appointment becomes a few minutes at the counter.

Trust, but verify

AI can sound confident while being wrong, especially when your data has gaps. Make it show its work before anything goes in front of a customer.

• Ask for the SKU or model number on every recommendation, so you can look it up in seconds.
• Check specs before you quote. Dimensions, lumens, voltage and ratings like damp or wet location are the details that cause returns.
• Confirm price and availability in your own system. Uploaded files may be out of date (see Step 3).
• Tell it to say “not in the catalog” rather than guess.

You can set this once as Project instructions so every chat follows it. Paste something like:

Only recommend products from the uploaded files. For every product, include the brand and SKU or model number. If a spec or price isn't in the files, say so instead of guessing. Don't search the web.

Keep it current: connect your product data

The fix for the snapshot problem is to stop uploading files and connect the AI to a source that updates itself. Claude supports connectors, which let it read from other tools live instead of from a file you uploaded last month. There are three ways a showroom can get there.

Option 1: a connected spreadsheet. Keep your products in a Google Sheet (or similar) and connect it. It’s better than re-uploading, but someone still has to keep the sheet accurate.

Option 2: your current website. If your site already holds your full catalog, it can be the AI’s source of truth. Whether that works depends on how your site is built and whether your provider exposes product data in a way AI tools can read. Most showroom sites weren’t built with this in mind. Our free AI readiness check scans your site and tells you where you stand.

Option 3: a site built for it. Neural Partners builds showroom websites with an AI connector included. Connect it to Claude once and your team can search your live catalog, with structured specs, current pricing and stock, from the same chat they already use. When a price changes on your site, the AI knows. Nobody re-uploads anything.

Whichever route you pick, the principle holds: the AI is only as good as the product data behind it. Get the data right and kept current, and Steps 1 through 4 just work.

Not sure where your site stands? Run the free AI readiness check. It takes a minute, and you’ll know whether your catalog is ready to connect or what’s in the way.