Stock count is a structured field. For vintage, almost every piece has a count of 1. SleekAI honors that, so a recommended item is genuinely available. Sold pieces are filtered out before the prompt reaches the model. The bot also adds urgency in the language ('the only one we have, want me to hold it') rather than treating every recommendation as if there were stock behind it for weeks.

Yes, if your sizing guide includes the mapping. Upload an era-by-era PDF (Italian 42 in 1995 sits closer to modern UK 10; US 6 in 1980 sits closer to modern US 8) into OpenAI Files. The bot retrieves the relevant row when a shopper asks for a 1990s piece in modern UK 8. Without that grounding the bot would default to modern sizing, which misleads.

Yes. Condition notes for each piece (excellent, very good with notes, fair with notes) feed into the prompt verbatim. The bot says 'lining replaced 2018, exterior pristine' or 'small underarm stain on left side, priced accordingly' rather than defaulting to 'beautiful condition.' Vintage retail trust is built on condition honesty, and the bot reinforces it.

Yes. Designer, year, country of origin, and any authentication notes feed into custom fields. A shopper asking 'is this real Halston or a licensed reissue' gets the right answer from the catalog rather than a guess. For pieces with paperwork (auction provenance, brand authentication card), the bot mentions what is included so the buyer knows.

Yes. When a shopper expresses serious interest in a one-of-one, the bot can route a hold request. Captures the piece, the customer's name and email, and the requested hold window. Staff sees the request in WordPress and confirms or declines. This matters for vintage because a 24-hour hold can be the difference between a sold piece and a lost one.

Your own. SleekAI is bring-your-own-key for OpenAI, Anthropic, Google, or OpenRouter. You stay in control of costs and customer data. For a vintage store with moderate monthly volume, a smaller model handles routine sizing and condition questions cheaply, with a larger model reserved for designer-deep conversations.

Yes. If your custom fields record alterations (hem shortened, sleeves taken in, lining replaced), the bot answers honestly. A shopper hearing 'the original hem was 2 inches longer, current length sits below the knee on a 5 foot 6 wearer' makes a better decision than one hearing a generic 'classic 1970s length' line. Alteration honesty also feeds the return conversation - the shopper cannot say 'I didn't know' if the bot told them.

Yes. Shoppers often ask 'are you sourcing any 1990s Comme des Garcons knits?' The bot captures sourcing interest and routes to the buyer's inbox. The buyer sees real requests with the customer's contact and can reach out when the next pickup happens to include a relevant piece. The conversation log becomes a sourcing wish list.