Kitchen and Furniture Stores: Booking, Not Baskets

A fitted kitchen does not go into a shopping basket. Between the first look at a website and the signature sit a design meeting, a technical survey, two or three revisions and a lead time counted in weeks. For furniture stores, upholstery specialists and fitted kitchen studios that means one thing: your website does not sell anything, it paves the way to appointments. The retail figures say the same. The home and furnishing sector as a whole reaches an online share of 21.6 percent (HDE Online-Monitor 2026) of its total market, while the institutional specialist retail trade in the very same sector reaches only 7.1 percent (HDE Online-Monitor 2026). That is not a weakness of the retailers, it is a property of the product: what has to be planned, measured and installed is not ordered at half past ten at night. Something else does get started at half past ten at night, though, namely the appointment, provided there is a way to request one. This article shows which questions an AI assistant in furnishing retail can answer reliably, which four details it has to collect before the appointment so the designer does not have to call twice, how display models are treated as one-off items, and where its limits are drawn on purpose.

Enlarge graphic

Labels in the illustration

• Questions before a showroom visit
• How long is the lead time?
• Is the survey charged extra?
• Is installation included?
• Is old furniture taken away?
• Is the display model still there?
• asked in the evening and at weekends
• 7.1
• percent online share in
• specialist retail 2025
• (HDE Online-Monitor 2026)
• Booking assistant
• Range and showroom
• Delivery and installation
• Design and survey
• trained on your sources
• ?
• Answer from the knowledge base
• with a date from service and delivery
• +
• Four details captured in chat
• room, size, timeframe, range
• Design appointment requested
• free advisory slots
• The assistant does not answer
• fixed prices
• binding dates
• right of withdrawal
• goes to the advisers
• When enquiries arrive
• example course of one day
• opening hours
• evening and night

AI assistant in the kitchen studio: appointment, not basket

The online channel here is an appointment channel

The official definition of the sector makes clear what is being discussed here. Home and furnishing includes, among others, economic activity 47.591 (Federal Statistical Office, cited in the HDE Online-Monitor 2026), retail sale of home furniture, which is exactly the kind of business with a showroom, a design service and an installation team. For the institutional specialist retail trade in that sector the study reports an online share of 7.1 percent (HDE Online-Monitor 2026) for 2025, after 7.0 percent the year before and 4.8 percent in 2020. The total market of the same sector, by contrast, sits at 21.6 percent (HDE Online-Monitor 2026), because pure online players and marketplaces count towards it. Measured against total online volume the sector accounts for 8.1 billion euros and therefore 8.8 percent (HDE Online-Monitor 2026). The gap between 21.6 and 7.1 is the real message: the sector's online revenue is largely generated elsewhere, while in the showroom it is still the conversation that sells.

The same study supplies the figure that slows down the obvious reflex: 40 percent (HDE Online-Monitor 2026) of bricks-and-mortar retailers sell goods over the internet, and the authors state explicitly that as a sales channel online is losing relevance for many retailers. A bigger shop system is therefore rarely the answer in furnishing retail. Concluding from this that the website is secondary, however, confuses the channel with the job. The website is where the decision to visit the showroom is made, and reading behaviour works against the retailer here: 79 percent (Nielsen Norman Group 1997) of users scan every new page instead of reading it. Lead times, survey fees, the scope of installation and disposal are somewhere in the body text, but not where somebody is looking for them at half past ten at night. The question stays open, and open questions rarely end in an appointment.

The questions that come before every design meeting

Anyone reading chat transcripts from furniture stores and kitchen studios finds little surprise and a great deal of repetition. Before the appointment, prospects want to know whether the trip to the showroom is worth it, how long the whole thing takes and what will be added at the end. They do not ask about design, they ask about process, deadlines and the costs beside the furniture. And they ask when they have time: after work, at the weekend, during a lunch break. The following topics come up again and again in furnishing retail projects.

• Lead time: how many weeks lie between order and delivery, and what the range depends on
• Technical survey: whether one is required, who carries it out and whether it is charged separately
• Installation: what assembly covers, who handles electrical and water connections and how long it takes
• Old furniture: whether the existing fittings are dismantled and disposed of, and at what cost
• Display models: whether a one-off item is still available, when it is removed and whether delivery is possible
• Design documents: what to bring to the appointment, whether a floor plan is enough and whether photos help
• Instalments: whether financing is offered and which partner handles it
• Directions and duration: where to park, how long a design meeting takes and whether children can come along

These questions can be answered without anyone stepping in, provided the answers sit cleanly and currently in a maintained source. That is exactly why the knowledge base is the foundation: the assistant is instructed to answer from approved building blocks, and sourced answers name their reference. Which content belongs in it and how it should be cut is covered in building a knowledge base for an AI chat. For a showroom that means something very concrete: the lead time range the assistant quotes is the range the sales team entered, and it is maintained in one place instead of five brochures. If the situation at the manufacturer changes, the answer in the chat changes the same day.

Four details the design team needs in advance

A design appointment without prior knowledge costs twice. The adviser calls to ask for dimensions, calls a second time because the timeframe does not work, and discovers during the meeting that the customer's expectation and the available range do not match. Four details prevent that in most cases, and all four can be collected in a chat without it feeling like an interrogation: which room is involved and what is meant to happen there, what the rough dimensions of that room are, when delivery should take place and what scale the project is on. The assistant asks them one after another and explains each one instead of presenting them as mandatory fields.

• Opening: the assistant asks about the project, so kitchen, bedroom, upholstery suite or fitted wardrobe, not for an article number
• Room and dimensions: rough length and width, layout, sloping ceilings, existing connections, optionally an uploaded floor plan
• Timeframe: the desired delivery or installation month plus the reason, so a move, a renovation or a replacement
• Scale: a range instead of a figure, introduced with a note on what the designer needs it for
• Existing items: whether appliances or furniture are to be kept and whether old furniture needs collecting
• Handover: the case reaches the design team in a structured form, as a note on the appointment and optionally, in the Custom package, in the merchandise system

The reason this reaches more prospects in a conversation than in a form next to it is friction. An average retail checkout still comprises around 11.3 form fields (Baymard Institute 2024), and around 22 percent (Baymard Institute) of online shoppers have abandoned a purchase because the process was too long or too complicated. A booking form with mandatory fields, a preferred-date picker and account creation is no better. How the same effect shows up in lead generation is described in qualifying leads via chat instead of a long form; the underlying service is the lead assistant, which passes the captured details on in a structured form. That the same pre-qualification works in other advice-heavy sectors is shown by the article on the AI assistant for real estate.

The design appointment starts in the chat

The second big point of friction sits behind the decision: finding a date. A callback request on Sunday evening is worked through on Monday, the callback reaches nobody, on Tuesday the showroom tries again, and on Wednesday the customer has an appointment somewhere else. A booking assistant instead offers free slots from the stored appointment types and records the chosen one as a request once the customer has agreed in the chat; the showroom confirms it in its overview. For the showroom little changes in the daily routine: the request arrives together with the note, the team confirms it, and the advisory calendar is maintained as before; a binding direct entry happens only through a set-up calendar integration (Custom package). The duration of the appointment type matters here: a kitchen design meeting is not a twenty-minute slot, and an appointment type that does not reflect that produces appointments nobody can keep.

For this to run reliably the assistant needs more than read access, it needs permission to carry out an action. The appointment mechanics in detail are described in appointment requests via a chat assistant, and the matching service is the booking assistant. A second operating mode is also worth having for the cases where a fixed slot comes too early: the assistant collects preferred times and the showroom confirms. Both variants save the usual chain of callback request, callback and date proposal. How strongly that affects the number of deals depends on location, range and staffing and cannot be quantified in general. What is solid is the observation that the stretch between first interest and first conversation is where prospects are lost most often.

Display models are one-off items

Furnishing retail has one product category that breaks every standard logic: the display model. It exists exactly once, it stands in exactly one showroom, it can only be delivered after the removal date, it has signs of use, and it is frequently the actual reason for the visit. An assistant that treats it like a stock item does damage: two prospects receive the same promise, and one of the two families drives out for nothing. The answer is not an availability indicator but a reservation request. The assistant records the interest with contact details and a deadline, marks the item as requested in the system and states explicitly that the showroom confirms the reservation. Only that confirmation is a commitment. The same restraint applies to the removal date: as long as it is not fixed, it is named as open rather than estimated.

The second particularity is the depth of advice required. Someone looking for an upholstery suite asks about cover material, seat firmness, depth and cleanability, and none of those sit in a price list. This is exactly where an assistant helps that does not list products but narrows them down: two or three follow-up questions about use, number of people and room size, then two suggestions with one sentence of reasoning each and the note that both are on display. The same mechanism is described in guided selling in the chat, only with a different goal: at the end there is no order but an appointment at which the right pieces are already waiting. That shortens the conversation in the showroom and raises the chance that the visit turns into an order.

What the assistant deliberately does not promise

In this sector an assistant that is allowed too much does more damage than one that can do too little. Three kinds of statement therefore have to be blocked. First, the fixed price for a planned interior: what a kitchen costs is decided by the design work, not by the chat. Second, the binding delivery date: it depends on the manufacturer, part deliveries and installation capacity and is quoted as a range or not at all. Third, any advice on the right of withdrawal. The reason is in the statute: under section 312g paragraph 2 number 1 (German Civil Code) there is no right of withdrawal for contracts on goods that are not prefabricated and for whose production an individual choice or specification by the consumer is decisive. Whether a specific design falls under that is a question of the individual case, and it is not decided in a chat. The classification of a contract concluded during a survey in someone's home does not belong in a chat either, even though section 312b paragraph 1 (German Civil Code) describes what an off-premises contract is. In both cases the assistant refers the matter to a person.

Technically this restraint rests on the assistant being instructed to answer exclusively from approved sources and to pass on anything uncertain instead of formulating whatever sounds plausible. How to secure that is described in preventing hallucinations with a clean knowledge base. Once a conversation reaches the limit, an orderly handover to human agents is the decisive step, and it matches what customers expect: 62 percent (Bitkom) of online shoppers want to turn to a quickly reachable human contact when there is a problem, while a chatbot is wanted by only 36 percent (Bitkom). For a purchase that will stand in the living room for ten years, that counts double. Price statements in a dialogue are a chapter of their own; which mandatory information has to travel with them is covered in what a chat assistant may say about prices.

Several showrooms, one answer

Furniture stores with several branches and studios with two or three showrooms have an additional problem: almost every answer depends on the location. Opening hours, advisers, display models, delivery area, installation partners and even lead times differ. An assistant that ignores this gives an answer that is right for the wrong branch in five out of ten cases. The answer is an early location context: the assistant derives it from the page being viewed, otherwise asks for it in one sentence, and keeps it for the whole conversation.

• Location first: derived from the page or clarified in one question, then fixed for the entire conversation
• Separate blocks: opening hours, directions and showroom areas exist per branch, not as one shared text
• Shared blocks: the design process, services and principles apply to every branch and are maintained once
• Appointment types per branch: each showroom gets its own appointment tool, and requests only take up its times, not those of a shared calendar
• Delivery area: the postcode decides which branch is responsible, and the assistant says so openly
• Holidays and special hours: deviating times are stored per branch and given an end date

How to set this up cleanly is covered in one assistant, many locations: answers per branch. In terms of data protection nothing special applies, only care: the details captured in the chat are contact data and information about somebody's living situation, so not the kind of thing that should sit in a shared mailbox without a deadline. Processing, storage location and deletion runs are described on the privacy and hosting page. In practice a short definition is enough: who sees the notes, how long they stay attached to the appointment, and what happens to a case that does not turn into an order.

Chat transcripts show where the description is unclear

The underrated benefit is not in the chat itself but in the analysis afterwards. If forty prospects ask about old furniture disposal in one week, it is missing from the services page. If the question about the technical survey always breaks off at the same point, the answer there is too vague. If a product line triggers many conversations but few appointments, something between expectation and showroom does not fit. These patterns are invisible without an assistant, because an unanswered question leaves no record. With one, they become the to-do list for the next revision of pages and brochures.

The route there runs through conversation analytics, which summarises topics, drop-off points and gaps; how to derive concrete improvements from that is shown in analysing and improving chats. It usually also becomes clear that part of the enquiries do not come from the person who buys: adult children research for their parents, couples split the preparatory work. How an assistant deals with relatives asking rather than the people concerned is described in the article on the AI assistant for home care providers. In furnishing retail it mainly means one thing: the assistant must not assume that the person asking is also the person planning.

A few steps to your booking assistant

Getting started is smaller than it sounds. It begins with a stocktake: which questions arrive today by email and phone, how much of that is already on the website, and what does only the person at the reception desk know? Out of that comes the first version of the knowledge base, usually from service pages, delivery terms, the installation description and a short collection of answers from the sales floor. After that the limits and handover routes are defined, the appointment types are stored with time windows and duration, the assistant is embedded with a short snippet and signed off with test cases. After two or three weeks the first analyses show which page should be sharpened. How appointment requests are taken in the chat, and why they only become binding once you confirm them, is shown by the booking assistant.

• Define lead time ranges per brand or product line and give each one a date
• Separate services cleanly: survey, installation, connection, disposal, delivery into the home
• Define the blocklist: no fixed price, no binding delivery date, no advice on the right of withdrawal
• Store appointment types per showroom and set the meeting duration realistically
• Define how display models are handled: a reservation request with a deadline instead of a promise
• Run test cases with price, date and legal questions and review the answers before go-live

Frequent questions