Ask for the things that change labour time: the type of clean, the property's size in bedrooms, bathrooms and floors (or floor area for offices), its current condition and when it was last cleaned, pets, extras such as ovens and inside windows, access and parking, and how often. With those answers an AI applies your pricing rules; without them it guesses.
Two questions decide most wrong cleaning quotes: service type and condition. A "three-bed house" can be a three-hour weekly tidy or a fifteen-labour-hour move-out after a tenant with a dog, and both customers describe it in the same four words. Customers also describe their own homes kindly, so the condition question needs a scale with descriptions and, for big jobs, a photo.
Group 1: what kind of clean this is
Ask this first. It's easy to answer, and it decides which pricing table every later answer feeds into.
| Question, as the chatbot asks it | Why it moves the price | How to verify |
|---|---|---|
| "Which clean do you need: regular (weekly or fortnightly), a one-off deep clean, a move-in or move-out clean, an after-builders clean, or a holiday-let changeover?" | Each has a different time per room. A move-out includes inside cupboards and appliances; a regular clean doesn't. | For move-outs, ask whether the property will be empty on the day |
| "Will the property be furnished or empty?" | Empty rooms clean faster but expose every mark on walls and floors | Confirm the date the keys are handed back |
| "Is there a checklist you need to meet, for example from a letting agent?" | Move-out checklists often require specific items (oven, windows, skirting) | Ask the customer to upload it |
Group 2: size, in the units your pricing uses
Ask for size in whatever units your price table is built on. For homes that's usually rooms; for offices, floor area. Don't ask for square metres from a homeowner who has no idea, and don't let the AI convert "a big three-bed" into a number.
| Question | Why it moves the price | How to verify |
|---|---|---|
| "How many bedrooms?" | Base time per bedroom | Cross-check against a property listing if they share a link |
| "How many bathrooms, and how many separate toilets or cloakrooms?" | A full bathroom takes far longer than a cloakroom; lumping them together under-prices | Photos for move-outs |
| "How many living or dining rooms? Is the kitchen open-plan?" | An open-plan kitchen-diner is one large room with a lot of surfaces | Photo of the kitchen |
| "How many floors, and are there stairs to clean?" | Stairs and landings add time and carrying | None needed |
Booking-form tools already work this way. BookingKoala, for example, lets you attach a price and a duration to each variable, such as the number of bedrooms and bathrooms, and even to "exclude" parameters that credit the customer for rooms they don't want cleaned. Your chatbot should follow the same logic: every answer maps to minutes and money in a table, and the AI's job is to collect the answers accurately, not to decide what they're worth.
Group 3: condition, the question customers answer too kindly
This is where AI quotes go wrong most often, so it gets the most care. Asking "how dirty is it?" gets "not too bad" from almost everyone. Give a scale with descriptions instead:
How would you describe the home right now?
1 - Cleaned professionally in the last month
2 - Kept tidy, no professional clean for a few months
3 - Lived in: some build-up in the kitchen and bathroom
4 - Heavy: grease on the hob or oven, limescale, pet hair throughout
5 - Very heavy: not cleaned for a long time, or after a party or tenancyThen ask the follow-ups that make the rating honest:
- "When was it last cleaned professionally?" A date is harder to shade than an adjective.
- "Any pets? How many, and do they shed?" Two long-haired dogs can add a third to the time on floors and soft furnishings.
- "Does anyone smoke indoors?" Nicotine film on walls and glass is slow to remove.
- "Any mould, heavy limescale or anything you're embarrassed about?" Asked kindly, this gets honest answers more often than you'd expect.
- "Could you send photos of the hob, the oven and the main bathroom?" For move-outs and anything rated 3 or above, make this required. Three photos settle most arguments before they start.
Group 4: extras that add fixed time
Extras are where profit leaks on one-off cleans, because customers assume they're included. List them as tick-boxes with their own time, which also makes them easy to sell.
| Extra | Illustrative added time | Question detail that matters |
|---|---|---|
| Inside the oven | 60-90 minutes | Single, double or range cooker |
| Inside the fridge or freezer | 20-40 minutes | Will it be empty and defrosted? |
| Inside windows | About 5 minutes per window | How many windows, and are any hard to reach? |
| Blinds | About 10 minutes each | Venetian slats take longest |
| Inside cupboards | 30-60 minutes for a kitchen | Empty or full? |
| Walls and skirting wiped down | Varies by room count | Usually expected on move-outs |
| Changing beds, laundry, ironing | Per bed or per load | Is fresh linen provided? |
The times above are illustrative. Use your own team's averages, and review them every few months.
Group 5: access, timing and the job-killers
These don't always change the price, but they decide whether the job can happen at all, and a missed one costs a wasted journey.
- "Will someone be home, or how will we get in?" Key safe codes and handovers go into the booking, never into the chat transcript that gets shared.
- "Where can we park, and is it free?" Paid parking or a long carry from the van adds cost; ask for it rather than discovering it.
- "Which floor is it on, and is there a lift?" Carrying equipment up five flights takes time.
- "Should we bring supplies and equipment, or use yours?" Affects price and what the team loads.
- "For move-outs: will the water and electricity be on?" A move-out with the power off can't be done properly.
- "Any alarms, pets at home or areas we shouldn't enter?"
- "How often, and which days or times suit?" Frequency changes the per-visit price, because the second weekly clean is lighter than the first.
Specialist cleans need their own follow-up questions
The five groups cover a normal home. Three kinds of one-off clean need extra questions, because the price depends on things the standard list never asks.
After-builders cleans. Ask what work was done (plastering and sanding produce fine dust that settles for days), whether there are paint or plaster splashes on glass and floors, whether stickers and protective film are still on new windows, and whether trades will still be working on the day. An illustrative answer, "new kitchen, walls replastered and sanded, film still on the bifolds", should push the quote into the heavy after-builders rate and add film removal as an extra. The same kitchen without plastering is a much lighter job.
Holiday-let changeovers. Ask the turnaround window between check-out and check-in, whether you supply linen and towels or the owner does, whether restocking (toilet roll, tea, dishwasher tablets) is part of the job, and how damage or missing items should be reported. A four-hour window for a three-bedroom let may need two cleaners rather than one, which changes the price even when the hours don't. Photos of any damage, sent to the owner before the next guest arrives, are usually expected.
End-of-tenancy cleans. Ask whether there's an inventory or check-in report, and ask the customer to upload it. The report shows the standard the property is judged against, and deposits are argued over it. Put a line in the quote about what a clean can't fix, such as worn carpet, burns or old stains, so the customer doesn't expect a clean to cover damage.
When answers are vague: the follow-ups the bot should ask
Customers answer quickly and loosely. Give the chatbot a follow-up for each vague answer, so it never has to guess:
| Vague answer | Follow-up the bot asks | Why it matters |
|---|---|---|
| "It's average" | Shows the 1-5 scale and asks them to pick | "Average" means a 2 to some people and a 4 to others |
| "A few windows" | "Roughly how many windows need doing inside? Are any above the ground floor stairs?" | Window time is per window; stairwell windows need a ladder |
| "Some pets" | "How many, and are they dogs, cats or something else? Do they go upstairs?" | Hair on stairs and beds adds time room by room |
| "Big kitchen" | "Is it open-plan with the dining or living area? Is there an island?" | An open-plan space is priced as more than one room |
| "Not too bad" | "When was it last cleaned by a professional?" | A date is more honest than an adjective |
Commercial cleaning needs a different list
If you also clean offices, don't run them through the domestic questions. An office quote depends on floor area, the number of desks, washroom cubicles and kitchens, floor types (carpet or hard floor), how often, whether cleaning must happen out of hours, security and key-holding arrangements, who supplies consumables such as paper and soap, and waste and recycling. Put commercial enquiries on a separate path in the chatbot, and have it collect those details and book a site visit rather than quote outright. Commercial prices usually depend on a walk-round and a specification, not a formula.
Turning answers into a price the AI can't improvise
A chatbot quotes accurately only if it applies rules you wrote. Here is an illustrative rule set, in the form you'd give an AI in its instructions:
PRICING RULES (apply exactly; never estimate outside them)
Move-out base minutes: bedroom 45, bathroom 60, cloakroom 20,
kitchen 90, living/dining room 30, hall and stairs 20.
Condition multiplier: 1 = 1.0, 2 = 1.1, 3 = 1.25, 4 = 1.5.
Rating 5 = do not quote; hand over to a person.
Pets: add 10% after the condition multiplier.
Extras (minutes): single oven 60, double oven 90, fridge 30,
inside windows 5 each, blinds 10 each.
Labour charge: $38 per labour hour. Minimum charge: $120.
Round the total up to the nearest $5.
Output: labour hours, team size, time on site, price, and the
assumptions the price depends on. Say the price is confirmed after
we've seen the photos.Worked through for an illustrative enquiry: a three-bedroom, two-bathroom house, move-out, condition 4, one dog, a double oven, inside the fridge, 12 inside windows and 4 blinds.
- Rooms: 3 x 45 + 2 x 60 + 90 + 2 x 30 + 20 = 425 minutes
- Condition 4: 425 x 1.5 = 637.5 minutes
- Pets: + 10% = 701 minutes
- Extras: 90 + 30 + 60 + 40 = 220 minutes
- Total: 921 minutes, about 15.4 labour hours, or just under 8 hours on site for two cleaners
- Price: 15.35 x $38 = $583, rounded up to $585
The same house as a fortnightly regular clean at condition 2 might be three labour hours and around $115. That's the gap a quote falls into when the service-type and condition questions are skipped. As with any AI arithmetic, check the first few dozen quotes against a spreadsheet using the same rules; the model can misapply a multiplier even with the rules in front of it.
A sample intake chat, and the follow-up that caught the truth
An illustrative exchange on a cleaning firm's website, trimmed:
Bot: Which clean do you need: regular, one-off deep, move-in or move-out, after builders, or a holiday-let changeover?
Customer: Moving out end of the month.
Bot: How many bedrooms and bathrooms, and any separate toilets?
Customer: 3 bed, 2 bath.
Bot: How would you describe it now, from 1 to 5? [scale shown]
Customer: Probably a 2, it's pretty clean.
Bot: Thanks. When was it last cleaned professionally, and are there any pets?
Customer: Never had a cleaner. We've got a dog.
Bot: No problem. For move-outs we ask for three photos (hob, oven, main bathroom) so the price is right first time. Could you add them here?
The photos showed a hob and oven closer to a 4 than a 2. The bot's rules said photos override the self-rating for kitchens and bathrooms, so the quote came out at the heavier rate with a line explaining why. Without the "never had a cleaner" follow-up and the photos, the quote would have been about $120 low, and the team would have run three labour hours over.
Two things to keep in how the bot talks. It should tell customers at the start that it's an automated assistant, and it should never promise what isn't in your services. Chatbot guardrails that stop AI promising what you don't offer covers the second in detail, and training a chatbot on your FAQs, policies and prices covers loading the rules.
Testing the questions against jobs you've already done
Before the chatbot quotes a real customer, run it on ten recent jobs where you know how long the work actually took. Answer the intake questions as that customer would have, and compare.
| Past job | Chatbot labour hours | Actual labour hours | Why the gap |
|---|---|---|---|
| 2-bed flat, move-out, rated 2 | 5.5 | 10 | Customer rated it kindly; no photos requested |
| 4-bed house, deep clean, rated 3 | 11 | 11.5 | Within tolerance |
| 3-bed, fortnightly regular | 3 | 3 | None |
| 1-bed, after builders | 6 | 9 | Plaster dust on every surface; after-builders multiplier too low |
In this illustration the test found two fixes: make photos compulsory for move-outs whatever the rating, and raise the after-builders time. That's the whole point of testing: the questions and rules improve before a customer pays for the gap. Testing a customer chatbot before it goes live has a fuller test plan, and once jobs are booked, photo checklists and quality control for cleaning teams closes the loop on whether the work matched the quote. For how the questions fit into the wider quoting and booking flow, see how cleaning companies use AI to quote and book jobs, and if you're still deciding between a phone assistant and a web chat, AI receptionist or chatbot for a cleaning company.
Keeping the quotes accurate after launch
Pricing rules drift as your team gets faster, products change and new kinds of work come in. Once a month, export the quotes the chatbot gave alongside the hours each job really took, and ask the AI to compare them by service type and condition rating. An illustrative summary: "Regular cleans within 5%. Move-outs rated 3 ran 18% over on average; move-outs rated 4 were within 6%. Oven extras ran 20 minutes over on range cookers." Two changes follow from that: raise the condition-3 multiplier for move-outs, and add a separate time for range cookers. Make one or two changes a month, not ten, so you can see which one worked.
Intake questions: what cleaning firms ask next
How many questions can I ask before people give up?
Fewer than you'd think if they're all asked at once, and more than you'd think if they arrive one at a time with a reason. A chatbot that asks eight to twelve short questions, skipping any that don't apply, usually feels quick. Put service type and size first, because they're easy, and leave photos until the customer can see the quote is nearly ready.
Should the chatbot give a firm price or a range?
Give a firm price for regular cleans of normal homes where the answers fit your rules, and a range for move-outs, after-builders cleans and anything rated heavy, confirmed once you've seen the photos. Whatever it gives, the chatbot should say what the price assumes, so a cleaner arriving to a different home has something to point to.
What if the customer's answers fall outside my pricing rules?
Tell the chatbot to stop quoting and hand over to a person whenever an answer falls outside the rules: more bedrooms than your table covers, a condition rating of five, hoarding, biohazards, or commercial premises on a domestic form. It should collect the details and a callback time, and say someone will be in touch, rather than inventing a price.
Further reads
- Can AI Help a Cleaning Company Rebook More Regular Clients? — Turn one-off quotes into regular clients once the first clean is done.
- How to Stop an AI Chatbot Giving Customers Wrong Answers — Keep the quoting bot from answering questions it shouldn't.
- How to Add an AI Chatbot That Captures Leads on Your Website — Place the intake chatbot on your site so it captures every enquiry.
- Can AI Price a Job Accurately? What Tradespeople Should Check — The checks any AI-produced price needs before a customer sees it.
- AI Chatbot Disclosure: What to Tell Customers at the Start of a Chat — What to tell customers at the start of an automated quote chat.
- How Dog Groomers Can Use AI to Collect Pet Details Before Booking — Collect breed, coat condition, temperament and health details before the booking, and let AI turn them into a pet card and suggested slot you approve.
- AI Video Survey Tools for Movers Compared: Cost and Accuracy (2026) — Six video survey tools for removal firms compared on published price, pricing model and what their accuracy claims actually measure.
- How Removal Firms Turn Customer Videos Into Quotes With AI — A seven-step workflow for removal firms: from the customer's phone video to an inventory, a checked volume and a priced quote with clear assumptions.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: BookingKoala help pages on pricing parameters and exclude parameters for booking forms. Checked September 2026.