A self-storage business can use an AI chatbot to recommend a unit size by feeding it a size guide with each unit's floor area and typical contents, then asking the customer what they are storing and how much. The bot adds up the items, adds access space, and suggests a size with a range. Capture the lead, then hand over to a person to confirm.
"What fits in a 10x10?" is the most common question a storage operator hears, and the honest answer is "it depends on what you own and how you pack". A chatbot can run that conversation at midnight, asking the same questions your best counter staff ask. This tutorial builds it step by step: the size guide that serves as the bot's knowledge, the item-to-size logic, the questions, the lead capture and the tests. All sizes and figures are illustrative, so replace them with your own units.
Step one: write the size guide the bot will answer from
The size guide is the whole of the bot's knowledge about your units, so it must be specific. Use your real unit mix. Below is a filled-in example for a hypothetical site, with metric equivalents. The "typical contents" column is the bit customers care about, and it is a rough guide.
| Unit | Floor area | Typical contents (illustrative) | Good for |
|---|---|---|---|
| 5x5 ft (1.5 x 1.5 m) | 25 sq ft | Boxes, a few small items, seasonal gear, a bike | Student storage, a walk-in-wardrobe's worth |
| 5x10 ft (1.5 x 3 m) | 50 sq ft | Contents of a small room: single bed, desk, boxes | Studio flat, a spare room |
| 10x10 ft (3 x 3 m) | 100 sq ft | Contents of a one-bedroom flat, or a bedroom plus boxes | Moving, short-term overflow |
| 10x15 ft (3 x 4.5 m) | 150 sq ft | A two-bedroom home's main furniture and boxes | House moves, downsizing |
| 10x20 ft (3 x 6 m) | 200 sq ft | Three-bedroom home's contents, or a small business stock | Full moves, trade stock |
| 10x30 ft (3 x 9 m) | 300 sq ft | A large home, or a vehicle plus contents | Large moves, vehicles, trade |
Add what the guide does not show on a table: door width and height, ceiling height, whether the unit is indoor or drive-up, climate-controlled options, floor level, lift access, and how access hours work. Then add a "do not store" list (your rental agreement decides this; a plain-English rental agreement draft using AI covers wording) and a plain statement of what you do not know: for example, that you cannot confirm a particular unit is free without checking the live system.
Capacity, not just area
Customers think in floor area, but stored goods stack upwards. A simple conversion helps the bot reason: usable volume is roughly floor area multiplied by a stacking height. With an assumed 8 ft ceiling and goods stacked to about 6 ft safely, a 10x10 gives a rough 600 cubic feet of usable space. Because packing is never perfect, the bot should treat a unit as about 70-80% usable once you allow for a walkway to reach items. These are illustrative assumptions; set your own from how your customers actually pack.
Step two: turn "what are you storing?" into an item list
People do not know their own volumes, so the bot should offer items to tick or describe, not ask for cubic feet. Give it a table of common items with a rough footprint, expressed as the floor area they take when stacked sensibly. These values are examples for the logic, not measurements of your stock.
| Item | Approx. floor area used (sq ft) |
|---|---|
| Double bed (frame and mattress stood up) | 10 |
| Three-seater sofa | 12 |
| Wardrobe | 6 |
| Dining table and four chairs | 14 |
| Washing machine or fridge | 4 each |
| Standard box (stacked 4 high) | 1 per 4 boxes |
| Bicycle | 3 |
The logic is then a simple sum. Add the items, add 20% for walkway space, and pick the smallest unit that covers the total. Spell this out in the bot's instructions so it does not improvise:
Sizing rule:
1. Sum the floor area of the customer's items using the item table.
2. Multiply by 1.2 for access space.
3. Recommend the smallest unit whose floor area is at least that number.
4. If the result is within 10% of the unit's area, also mention the next size up.
5. Show your sums briefly and say the result is an estimate.
6. If the customer lists items not in the table, ask for width, depth and
height, or say you will pass it to the team.A worked example with numbers
A customer says they are moving out of a one-bedroom flat for three months and have a double bed, a three-seater sofa, a wardrobe, a small dining table with four chairs, a washing machine and about 24 boxes.
| Item | Floor area (sq ft) |
|---|---|
| Double bed | 10 |
| Three-seater sofa | 12 |
| Wardrobe | 6 |
| Dining table and four chairs | 14 |
| Washing machine | 4 |
| 24 boxes, stacked 4 high (24 / 4) | 6 |
| Items total | 52 |
| Plus 20% for access (52 x 1.2) | 62.4 |
62.4 sq ft falls between the 5x10 (50) and the 10x10 (100). The smallest unit that covers it is the 10x10, so the bot recommends that, and because 62.4 is well under 100 it can add that there is spare room for a few extra boxes. If the customer later mentions a second sofa and a desk, the sum rises to about 85 and the 10x10 remains a snug fit. This is why showing the sums matters: the customer sees why, and can correct a missing item.
The illustration also shows a weakness: floor-area logic ignores height. A customer storing tall items that cannot be stacked may need more space than the sum suggests, so keep a human check for unusual contents.
Step three: design the conversation
Keep it short. A reasonable flow has five questions, asked one at a time:
- What are you storing, in your own words? (Free text, or a picker for common items.)
- Roughly how many boxes? (A range is fine.)
- For how long, and from what date? (This is also a lead-quality signal.)
- Anything large, fragile or unusual? (A piano, a motorbike, a tall display cabinet.)
- Will you need to visit often? (Frequent access favours drive-up or ground floor, and shapes the walkway you leave.)
Then the bot gives a recommendation with its sums, a one-line caveat, and an offer: reserve online, book a viewing or leave details for a call back. A short conversation, with a sample reply (illustrative):
Customer: I have a one-bed flat to store for 3 months. Is a 10x10 enough?
Bot: A 10x10 (100 sq ft) usually holds a one-bedroom flat's contents. To check
against your things, could you tell me the large items you have: bed size, sofa,
wardrobe, table, white goods, and roughly how many boxes?
Customer: Double bed, 3-seater sofa, wardrobe, small dining table, washing machine, 24 boxes.
Bot: Adding those up gives about 52 sq ft of items, or about 62 sq ft with space to
walk in. A 10x10 fits comfortably with room for a few more boxes. A 5x10 would
be too tight. When do you need it from, and may I take a name and phone number
so the team can confirm availability and today's price?Notice that it does not claim the unit is free or give a price. Both come from your management system, and a chatbot that cannot see live availability must not pretend to. For the wider question of whether a bot should book or only advise, can a hotel chatbot take bookings or only answer questions applies almost word for word.
Step four: instructions that keep the bot inside its lane
Write the bot's instructions in plain words. A working starting point:
You are the sizing assistant for [site name]. You help people choose a storage
unit size. Use ONLY the size guide, item table and policies provided.
- Show the sums and say the result is an estimate.
- Never state a price or say a unit is available; say the team will confirm.
- If something is not in your sources, say so and offer a call back.
- Do not give advice about insurance, legal matters or what is safe to store.
Point to the rental terms and the team.
- Tell the customer you are an AI assistant in your first message.
- Ask for name and phone number or email only after giving a recommendation.The same sheet logic as in other chatbot launches applies: one source of truth, a refusal list and a hand-off route. Building the FAQ a chatbot needs before launch gives the template for the question-and-answer half, which for storage covers access hours, gate codes, payment dates and what happens if someone falls behind.
Step five: capture the lead without scaring the customer
The size recommendation is the reason a visitor talks to the bot, and the lead is why you built it. The best moment to ask for details is right after the recommendation, when the customer has had value. Ask for the minimum: name, a phone or email, move-in date and the recommended size, which the bot already knows. Add a plain line on how you will use the details, and a link to your privacy notice.
Send each lead to a place your team will see it within minutes: an email, a CRM entry or a row in a spreadsheet, with the full conversation attached. A filled-in lead record (illustrative):
Name: [first name]
Contact: [phone or email]
Move-in: 14 days from today
Duration: about 3 months
Recommendation: 10x10 (items 52 sq ft, 62 with access); mentioned 5x10 too tight
Notes: has a washing machine, wants ground-floor access
Source: website chat, 9.40 pmTiming is the competitive edge. A lead logged at 9.40 pm and answered at 8.30 the next morning beats one that waited until Monday. For the phone side of the same problem, whether an AI call centre can rent a unit after hours weighs voice against chat, and Smith.ai's AI Receptionist, which was listed in September 2026 as free for 25 calls and then $150 for 75, is one of the options there.
Three other customers the bot meets, with their sums
Household moves are only part of the traffic. The bot's logic should cope with at least three other types, each with its own twist. Figures are illustrative.
A student with a term's worth of belongings. A few suitcases, a bicycle, a small fridge and 10 boxes. Sum: suitcases 4 (two at 2 each), bicycle 3, fridge 4, boxes 10 / 4 = 2.5, total 13.5; with 20% access, 16.2 sq ft. That fits the 5x5 (25 sq ft) with room to spare. The bot's reply should say a 5x5 is enough, and mention that it is the smallest unit so no cheaper size exists. For a student the real question is often the date, since terms end together and units fill up, so the bot should ask for the move-in date early.
A small trader storing stock. Say a market seller with 40 plastic crates, two folding tables and a gazebo. Crates stack, but the trader will be in and out weekly, so the walkway matters more than for a household. If crates are 1 sq ft each and stack three high, 40 crates is about 13.3 sq ft, plus tables 4, gazebo 3, total about 20, or 24 with the standard 20% allowance. Because access will be frequent, the bot should add a bigger allowance (say 40%) and mention ground-floor or drive-up units: 28 sq ft points to the 5x10, with the 5x5 too tight. Put this rule in the instructions: "for frequent access, use 1.4 instead of 1.2". The trade customer is also your most valuable one, so the bot should flag them as a business lead for a person to call.
A homeowner mid-renovation. Furniture from two rooms, plus tools and building materials. The materials are the problem: long lengths of timber and boards do not fit a floor-area sum. Instruct the bot to ask for the longest item and to check it against the unit's depth and door width. If the longest item is 3.6 m (about 12 ft), a 10x10 will not take it flat, and the bot should suggest a 10x15 or a human check. A single question about the longest item catches most of these cases.
When the customer cannot describe what they have
Some people simply do not know. They are clearing a parent's house, or they have a garage full of things they have not looked at in years. Give the bot two exits so it does not guess:
- A rule of thumb by home size. Offer the "typical contents" column from the size guide: "A one-bedroom flat's contents usually go in a 10x10. Does that sound right?" It gives a starting point without pretending to know.
- A photo or a visit. Offer to have the team look at a photo of the room or garage, or to book a viewing of a unit of the suggested size so they can see how much space it is. A person judging a photo is more reliable than a model's guess, and it also produces a warm lead.
A useful instruction is: "If the customer cannot list items, do not estimate from guesses. Offer the typical-contents guide and a call back." This stops the bot from making up an inventory to be helpful.
Keeping the size guide honest over time
The bot is only as good as the guide, and guides drift. Units are converted, unit mixes change, a promotion makes a size scarce, and the walkway allowance you set may be wrong for your customers. Review the guide after the first month, then each quarter, using three sources: the conversations where staff changed the recommended size, the questions the bot could not answer, and the units that sit empty longest. If staff keep upgrading customers from the 5x10 to the 10x10, your item table is too optimistic and the allowance too low. If the 10x15s sit empty while the bot sends everyone to 10x10s and 10x20s, check that the 10x15 appears in the guide with sensible typical contents. Keep a dated change log so that when a recommendation looks wrong you can see what changed.
Choosing a chatbot builder
You do not need a custom build. Most small chatbot products let you upload a document, set instructions and embed a widget. From the vendors' pages in September 2026: Chatbase's free plan gives 50 message credits a month (free agents are deleted after 14 days of inactivity), Tidio gives 50 lifetime AI conversations on free, and Botpress gives 25 conversations a month. These are test allowances, enough to prove the idea in a week but not to run a busy site. Check the current pricing pages, because vendors change them often.
When comparing, ask whether you can restrict answers to your own documents, view conversation logs, collect leads and send them to your email or CRM, and switch the bot off quickly. Ignore the length of the feature list.
Testing before customers see it
Write twenty test conversations from real enquiries and run them all. These example cases cover the usual weak spots:
| Test | What a good bot does |
|---|---|
| "What fits in a 10x10?" | Gives typical contents, then asks what they are storing |
| A full list of items for a two-bedroom house | Shows the sum, recommends 10x15 or 10x20 with a range |
| "I have a motorbike and some tools." | Asks dimensions, mentions drive-up or ground floor, hands to the team |
| "How much is the 10x10?" | No invented price; offers the link or a call back |
| "Is a 10x10 free this Saturday?" | Says the team confirms availability |
| "Can I store paint, gas bottles or food?" | Points to the prohibited list and the rental terms |
| "Will my things be insured?" | Does not advise; points to the terms and the team |
| A very short answer: "Just boxes." | Asks how many, then gives a range |
| A joke or rude message | Stays polite, returns to the task |
Also test the lead capture: submit a conversation and confirm the lead arrives with the transcript, in the right inbox, within the time you expect. Then ask a colleague who has never seen the bot to try it and watch where they hesitate.
Where unit-size bots go wrong
Overconfident sizes. A confident "a 5x10 will do" for a customer who underestimated their boxes leads to a second trip and an upgrade conversation on day one. Fix: the 20% access allowance, the "within 10%, mention the next size" rule and a range in every answer.
Quoting from memory. A model may recall typical storage sizes from its training and mix them with yours, so a size you do not offer turns up in the answer. Fix: the instruction to use only provided sources, and a test that asks about a size you do not have.
Dropping the lead. A good chat that ends without capturing details is wasted. Fix: a gentle request after the recommendation, and a visible phone number for people who prefer to talk.
Hiding that it is a bot. Tell customers up front. If you have customers in the EU, transparency rules for chatbots apply, and a short disclosure is good practice everywhere; how to write an AI disclosure statement shows the wording.
Measuring whether it works
After a month, count conversations, leads captured, how many leads became reservations, and how many recommended sizes the team changed after a call. The last number is the most useful: if staff change the size in a large share of cases, correct the item table or the walkway allowance. Read ten conversations a week for the first month, fix the guide rather than the prompt, and add any item customers keep mentioning that is missing from your table. For the broader view of where AI helps along the customer path, mapping your customer journey to find where AI helps is a sensible next read, and the AI guide lists the rest of the series.
Questions operators ask about unit-size chatbots
Should the chatbot quote prices?
Only from a live rate table or a price list you maintain in the bot's source. Rates change with occupancy, climate control and promotions, so a stale price causes disputes at the counter. If you cannot feed it live rates, have it recommend a size, then link to your reservation page for the current price.
How accurate can an AI size recommendation be?
Reasonable for a typical household, rough for odd or bulky items, and never exact. Customers describe contents loosely. Treat the answer as a starting size with a range, say 'a 10x10, or a 10x15 if you have more than the list', and give a tour or photo check for borderline cases.
Do I need a developer to build this?
Not for a first version. Most chatbot builders let you paste a size guide and instructions, then embed a script on your website. The harder work is writing the size guide and the logic. A developer helps only if you want the bot to write straight into your management software.
What should the bot do with the customer's details?
Ask for the minimum: name, phone or email, move-in date and the recommended size, with a clear note on how you will use it. Store leads in your CRM or a spreadsheet, tell people they are talking to an AI, and check your data-protection duties with a qualified adviser.
Further reads
- How to Add an AI Chatbot That Captures Leads on Your Website — Add a lead-capturing chatbot to your site.
- Can AI Qualify Wedding Leads by Date, Budget and Guest Count? — Qualifying leads by date, budget and size.
- What to Do When an AI Receptionist Gets a Booking Wrong — Fixing mistakes when automation gets it wrong.
- AI Chatbot for CRM Systems: Send Website Enquiries to Job Software — Send chat enquiries into job or CRM software.
- AI for a Hardware Store: Can It Answer "What Do I Need for This Job?" — How to test an AI assistant on real job questions at the counter, build the product-knowledge sheet it answers from, and where staff must step in.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: facts.md baseline for chatbot free tiers (Chatbase, Tidio, Botpress) and Smith.ai AI Receptionist pricing, September 2026; unit sizes in this tutorial are illustrative and should be replaced with your own.