A customer is twenty minutes away and about to get in the car. Before they do, they check two things on their phone: are you open, and have you got it.
Get both answers in front of them and you get the visit. Leave either unanswered and there is a good chance they go somewhere that told them, or order it online instead. That is the entire funnel for most physical retail, and it collapses into about fifteen seconds of website use.
Opening hours are not a trivial question
It sounds too basic to matter. It is the most-asked question in local retail by a wide margin, and a surprising number of shops answer it inconsistently across their website, their Google listing, and their door.
Bank holidays are where it falls apart. Every retailer has a story about someone who drove over on Easter Monday because the site said nothing. A bot answering from a single maintained source at least means the answer is consistent, and it can handle the phrasing customers actually use: "are you open now", "what time do you shut", "are you open Sunday".
Keep the underlying page current. A bot repeating last year's Christmas hours is worse than no bot, because it sounds authoritative.
Stock, and the honesty problem
"Have you got it in" is the question with the highest commercial value and the highest risk of getting wrong.
If your inventory is genuinely on your website, a bot can answer from it and that is a strong advantage over every competitor whose site is a brochure. If it is not, do not let the bot improvise. Configure it to explain that stock moves quickly, capture the item and the customer's number, and have someone check and call back.
That second version still beats silence comfortably. A customer who gets a callback within the hour saying "yes, I've put one aside for you" is a customer who arrives, and often buys more than the one thing.
What must not happen is a confident "yes, we have that" from a model reasoning about a product page. A wasted journey costs you the sale and the relationship.
Reservations and collection
Click-and-collect changed what people expect from local shops, including shops that do not formally offer it.
A bot can take the enquiry regardless: what the customer wants, whether they want it held, when they will come in, and a contact number. Holding an item for a named person for twenty-four hours costs nothing and converts at a rate that surprises retailers who have not tried it.
The same mechanism handles special orders. Someone wanting a size you do not carry, a colour you can get in, or a discontinued part is a real sale that currently walks away. Capturing it at the moment of interest, with enough detail to check with a supplier, turns a shrug into an order.
Product questions belong here too
Retail staff answer the same product questions endlessly, and a good number of them are asked by people who never make it into the shop.
Dimensions, materials, compatibility, whether something needs assembly, whether a part fits an older model, what the warranty covers, whether you take the old one away. All documented, all sitting somewhere a customer cannot easily find.
Answering these well does something beyond convenience: it establishes that you know your products, which is the main reason anyone chooses a specialist shop over a marketplace. That is worth surfacing rather than burying.
The awkward ones
Price matching is asked constantly and answered vaguely. Publish the policy, including its limits, and let the bot state it. Ambiguity here wastes staff time at the till and produces arguments in front of other customers.
Returns are the same. What the window is, whether a receipt is needed, what happens with sale items, whether you refund or exchange. Consistent published answers reduce the friction that makes people buy from a marketplace instead.
Where to stop it
No stock promises it cannot verify. No pricing that is not published. No commitments about delivery dates that depend on a supplier. No advice on whether a product is safe for a particular use, which for anything electrical, structural, or child-related should go to a person.
And it should not attempt to handle a complaint. Someone unhappy about a purchase wants a human, quickly.
Getting it in place
Point it at your product pages, hours, location, returns policy, and any stock feed you publish. If your opening hours live in three places, consolidate them first. The exercise of getting the content straight tends to fix problems that predate any chatbot.
Set the capture fields to item, timeframe, and contact, so a stock enquiry your bot cannot answer still becomes a callback.
Reading the result
The number to watch is enquiries captured outside trading hours, particularly evenings and Sundays, which is when people plan their shopping.
Then look at how many became a visit or a held item. Local retail lives on footfall, and the gap between "thought about coming in" and "came in" is usually one unanswered question.