Pick a chat tool that connects to your CRM or inbox, load it with your prices, policies and FAQs, decide the three or four details that make a lead, and have the bot ask for them only after it has answered the visitor's question. Paste its snippet into your site, then test it with twenty real questions before going live.
Budget anything from nothing to about $100 a month for a small site's traffic. HubSpot's free chatbot builder handles scripted lead capture at no cost; AI-written answers cost extra, either as a subscription add-on or per conversation. The order of the conversation matters more than the brand of tool: a bot that demands an email address before it helps gets closed, while a bot that helps first earns the details.
Answer first, then ask: the order that earns contact details
Most lead-capture chatbots fail in the first ten seconds. They open with "Hi! Can I get your name and email?" and the visitor, who came to find out a price or whether you have something in stock, closes the window. The fix is a simple rule you'll build into every stage below: the bot gives something useful first, and asks for contact details at the moment the visitor wants something only a person can provide, such as a firm quote, a reservation, a call back or a document.
Here's the difference on a self-storage website, where most visitors want a price and availability.
Before (form first): "Welcome! Before we start, please enter your name, email and phone number." Visitors who wanted a price in five seconds now face a form in a chat window. Many leave, and those who fill it in often type junk to get past it.
After (help first): Visitor: "How much is a small unit for about three months?" Bot: "Our Small unit (roughly the size of a large wardrobe, good for boxes and a bike) is $89 a month, and the first month is half price at the moment. We have four free at the main site. Would you like me to hold one for you until Friday? I'd just need your name and the best number to confirm it."
The second version asks for less than the form did, yet collects more real phone numbers, because the visitor now has a reason to give one. Everything below is about making that pattern reliable.
Write down what a lead is for your business
Before choosing a tool, write a lead card: the handful of details your team needs to act on an enquiry without phoning back to ask basic questions. Keep required fields to three or four. Every extra question costs you some visitors, so each one has to earn its place.
Here's a filled-in lead card for a storage facility. Use it as a pattern for your own business.
| Field | Why the team needs it | How the bot asks | Required? |
|---|---|---|---|
| What they're storing | Decides the unit size and whether they need a drive-up unit | "What are you planning to store? A rough list is fine." | Yes |
| Move-in date | Separates this week's customers from next season's browsers | "When would you like to move in?" | Yes |
| Length of stay | Long stays qualify for the prepay discount | "Roughly how long do you think you'll need it?" | No |
| Name | To hold a unit and greet them on the phone | "Who should I put the reservation under?" | Yes |
| Phone or email | To confirm the booking | "What's the best number or email to confirm it?" | Yes (either) |
| Permission for follow-up | Lets you send a reminder if they don't book | "Can we send you one reminder if the unit is still free next week?" | No |
Add one qualification rule to the card: which answers make a lead "hot". For the storage facility, a move-in date within seven days triggers an immediate text to the manager's phone; anything later goes into the CRM for the next morning. A car dealership might mark a lead hot when the visitor names a specific vehicle and has a part-exchange; a wholesaler might mark it hot when the visitor gives a trade account number. One rule is enough to start with.
Four realistic ways to add the chatbot
You don't need the most capable bot on the market. You need one that answers from your own information, collects the lead card fields, and pushes them into the place your team already checks. These four cover most small businesses; prices are list prices as of September 2026 and change often, so confirm them before you buy.
| Tool | How answers work | Lead capture | Price to start | Best fit |
|---|---|---|---|---|
| HubSpot chatbot builder | Scripted flows on the free plan; AI Customer Agent needs a Professional or Enterprise plan | Creates contacts in HubSpot's CRM automatically; can offer a meetings link | Free for scripted bots; AI agent about $0.50 per resolved conversation (50 HubSpot credits) | Businesses already on HubSpot, or happy to start its free CRM |
| Tidio | Lyro AI agent answers from your content; Flows handle scripted steps | Flows collect data and quote requests | Free plan; Starter from $29 a month (less billed annually); Lyro add-on from about $39 a month for 50 AI conversations | Small sites that want AI answers without a big platform |
| Intercom | Fin AI agent answers from your help content | Collects details through workflows and hands off to your team | Seats from $29 a month (annual); Fin from $0.99 per outcome | Businesses expecting heavier support volume alongside sales |
| Crisp | AI chatbot on the Essentials plan and above | Workflow builder collects contact fields | $95 a month per workspace, including about $25 of AI credits (roughly 450 automated conversations) | Teams wanting chat, email and social messages in one inbox |
Two notes on those prices. Per-outcome pricing (Intercom's Fin, HubSpot's Customer Agent) is cheap when volume is low and gets expensive if the bot handles hundreds of chats, so estimate your monthly conversations first. And subscription add-ons like Lyro are sold in conversation bundles, so a busy month can push you into the next tier. You can see the current figures on HubSpot's chatbot builder page, Tidio's pricing page and Intercom's pricing page. If you're weighing up a vendor that isn't listed here, the questions to ask a chatbot vendor before signing will save you a bad contract.
If your website builder or CRM already includes a chat widget, test that first. The best lead-capture bot is often the one that writes straight into the system your team opens every morning, even if its AI is less clever.
Setting it up, stage by stage
Allow a working day spread over a week. The timings below assume one person who knows the business well and has admin access to the website.
Stage 1: a one-page knowledge sheet (about 2 hours)
The bot can only answer from what you give it. Write one plain document with your prices, what's included, opening hours, access rules, payment terms and the twenty questions customers ask most. Write it in the words customers use ("Can I get to my stuff on Sunday?"), not your internal jargon. The fuller version of this job is covered in building the FAQ your AI chatbot needs before launch; for lead capture, the price and availability sections matter most.
An excerpt from the storage facility's sheet:
UNIT SIZES AND PRICES (monthly, standard rate)
Small - size of a large wardrobe; boxes, bike, small furniture - $89
Medium - size of a single garage half-full; contents of a 1-bed flat - $149
Large - size of a single garage; contents of a 3-bed house - $229
Offer: first month half price on stays of 3 months or more.
Prices above are the only prices. For anything else say:
"I'll ask the manager to confirm that price for you."
ACCESS
Gate open 6am-10pm every day. Office staffed 8.30am-5.30pm Mon-Sat.
WHAT WE DON'T STORE
Food, fuel, gas bottles, anything alive, anything illegal.Notice the instruction inside the sheet about prices that aren't listed. That single line prevents the most expensive mistake these bots make.
Stage 2: the bot's instructions (30 minutes)
Most tools have a box for the bot's behaviour or persona. Paste in something like this and edit it for your business:
You answer questions for [business name], a self-storage facility.
Answer only from the knowledge sheet. If the answer isn't there,
say you'll check with the team and offer to take their details.
Order of the conversation:
1. Answer the visitor's question fully and briefly (under 60 words).
2. If they show buying interest (asking about price, availability,
move-in, reserving), offer the next step: hold a unit, a call back,
or a quote.
3. Only then ask for: what they're storing, move-in date, name,
and phone or email. Ask one question per message.
4. Confirm what happens next and when a person will contact them.
Office hours are 8.30am-5.30pm Mon-Sat. Outside those hours,
say "by 9am the next working day".
Never invent prices, discounts or availability.
Never ask for card details or ID numbers in chat.
If someone is upset or asks for a person, hand over immediately.Then run a test chat and read the output critically. A typical first result (illustrative):
Visitor: Do you have anything big enough for a sofa and a fridge?
Bot: Absolutely! Our Medium unit would be perfect for a sofa and fridge, and we'd be delighted to help you with your storage journey. It's $149 a month. To get started, could you please provide your name, email address, phone number and preferred move-in date?
Three things to fix. "Storage journey" is filler, so tell the bot to avoid marketing language. It asked for four details in one message, which reads like a form, so repeat the one-question-per-message rule more firmly. And it didn't mention the half-price offer, which is the best reason to reserve; add "mention the current offer whenever you give a price" to the instructions. After those edits the reply becomes: "A Medium unit fits a sofa and a fridge comfortably. It's $149 a month, and the first month is half price if you stay three months or more. Would you like me to hold one?"
Stage 3: decide exactly when the bot asks
Write the triggers down so they don't depend on the AI's mood. For most businesses these four are enough:
- The visitor asks about price, availability or delivery dates for something specific.
- The visitor asks a question the knowledge sheet can't answer.
- The visitor asks for a person, a call or a quote.
- The conversation has gone five messages without a next step.
The second trigger is where many leads hide. "Can you store a classic car that doesn't run?" isn't on the sheet, so the bot shouldn't guess; it should say it'll check and take the details. For more on getting those handovers right, see when a chatbot should hand over to a human.
Stage 4: where each lead goes (about 1 hour)
A lead that sits in the chat tool's own inbox until someone remembers to log in is barely a lead. Connect the bot so that every completed lead card does three things: creates or updates a contact in your CRM, sends a summary to a shared inbox, and, for hot leads, pings a phone. HubSpot does the first natively; Tidio, Intercom and Crisp connect to most CRMs directly or through Zapier or Make.
Ask for the summary in a fixed layout so it can be read in five seconds:
NEW CHAT LEAD - HOT (move-in within 7 days)
Name: [name] Phone: [number] Email: [email]
Storing: sofa, fridge, 15 boxes
Suggested unit: Medium ($149, half-price first month offered)
Move-in: this Saturday Stay: about 4 months
Asked about: weekend access, van parking
Promised: call back by 9am Friday
Transcript: [link]The "Promised" line matters. It tells the person picking up the lead what the bot has already committed you to, so nobody is caught out.
Stage 5: install and position the widget (30 minutes)
Paste the snippet into your site's header settings or install the tool's plugin. Then decide where the bot appears. On the storage site, the prompt ("Need a price for your stuff? Ask me") shows on the prices, unit sizes and contact pages after 20 seconds. It stays quiet on the moving-tips guides and the careers page. Check it on a phone: on small screens the chat bubble often covers a "Book now" button or the cookie banner, and fixing that position takes two minutes once you've spotted it.
Stage 6: a twenty-question test before launch (1 to 2 hours)
Write twenty test questions from real enquiries (old emails and call notes are a goldmine) and run them all. Grade each answer pass or fail against the knowledge sheet. The full pre-launch testing routine goes further; the minimum is this mix:
| Test type | Example question | Pass looks like |
|---|---|---|
| Price on the sheet (6 questions) | "How much is the large one?" | Exact sheet price plus the current offer |
| Price not on the sheet (3) | "What do you charge for a wine collection in a climate-controlled unit?" | No figure given; offers to check and takes details |
| Buying signal (4) | "Can I move in tomorrow?" | Answers, then offers to hold a unit and asks one question at a time |
| Awkward or off-topic (3) | "Can I store fireworks?" | Clear no, from the "don't store" list |
| Wants a person (2) | "I just want to talk to someone" | Immediate handover with opening hours |
| Out of hours (2) | Asked at 11pm on a Sunday | Honest reply time, no promise of an instant call |
Don't launch with any failure in the "price not on the sheet" row. That's the one that turns into an argument at the counter.
A storage facility's first month with the bot
To see how the numbers can come together, take an illustrative single-site storage facility with about 2,100 website visits a month and a contact form that produced around nine enquiries a month. The manager set up Tidio Starter with the Lyro add-on, following the stages above, in about six hours across a week.
| First month | Figure |
|---|---|
| Chats opened | 95 |
| Real questions (not "hi" and gone) | 61 |
| Lead cards completed | 34 (56% of real conversations) |
| Of those, hot (move-in within 7 days) | 12 |
| Units booked from chat leads | 14 |
| Contact-form enquiries in the same month | 6 |
| Wrong answers found in weekly review | 3 (two access-hour mix-ups, one unlisted price) |
The quick sum: tool cost about $68 a month billed monthly, plus roughly 30 minutes a week reviewing transcripts. If an average booking is worth $110 a month for five months, each booking is worth around $550 in revenue. Even if half of those fourteen customers would have booked anyway through the form or the phone, seven extra bookings is about $3,850 of revenue for a $68 tool. Be strict with yourself here, though: count a chat lead as new only if that person hadn't already emailed or called. The form dropping from nine enquiries to six is a sign that some people simply switched channel. The deeper version of this sum is in what a website chatbot costs and whether it pays off.
The three wrong answers are the other lesson. Two came from the gate hours being described differently on two pages of the website, which the bot had also read. The fix was to correct the website, not the bot.
Where lead-capture bots go wrong, and how you'd notice
These are the failures I'd look for in the first month, each with the signal that gives it away.
- It invents a price. A visitor asks about a size or service you don't list, and the bot produces a confident figure. You'll spot it in transcripts, or when a customer quotes it back at the counter. Fix: the "only these prices" line in the knowledge sheet, plus a weekly search of transcripts for dollar signs. There's more on this in stopping a chatbot giving customers wrong answers.
- It asks for details too early. Your chat analytics show lots of conversations ending after the bot's first or second message. Fix: move the lead questions behind the triggers in Stage 3.
- Leads land where nobody looks. A courier firm's bot captured a Saturday request for an urgent Monday collection, but the alert went to an inbox nobody opened until Tuesday. You'll notice it as leads with no reply logged. Fix: send hot leads to a phone, and add "check chat leads" to the opening routine.
- It promises what you can't deliver. "Someone will call you in 10 minutes" typed by the bot at 11pm. Fix: put your real reply times in the instructions and test out of hours.
- It collects a number without permission to use it. The visitor gave a phone number to confirm one booking, and it ends up on your marketing text list. Fix: a separate, optional permission question, stored as its own field.
- It chats endlessly with time-wasters. Competitors fishing for prices or people testing the AI run up per-conversation charges. Fix: cap conversations at around fifteen messages and hand over or close politely.
Telling visitors it's a bot, and what you may keep
Say plainly that visitors are talking to an automated assistant, in the greeting and in the widget's name ("Storage assistant (automated)"). It's good manners, it lowers expectations about nuance, and if you sell to customers in the EU, the EU AI Act's Article 50 transparency duty (applying since 2 August 2026) requires people to be told when they're interacting with an AI system unless it's obvious. Link your privacy notice from the widget, collect only the lead card fields, and never ask for card numbers, ID documents or health details in chat. Check how long the chat tool keeps transcripts and set a retention period that matches your privacy notice; if you're unsure what your data-protection duties are for chat transcripts, ask your data-protection adviser.
Four numbers to check every Friday
Spend twenty minutes a week on these, and keep a simple log. Here's a filled-in example from the storage facility's third week:
| Number | How to get it | Week 3 | What would worry me |
|---|---|---|---|
| Capture rate | Lead cards completed ÷ real conversations | 58% | Under 30%: the bot is asking too early or answering badly |
| Hot-lead reply time | Time from alert to first human contact | 41 minutes | Over half a day |
| Lead-to-booking rate | Bookings ÷ lead cards (from the CRM) | 38% | Falling for three weeks running: check lead quality |
| Wrong answers | Read 10 random transcripts | 1 of 10 | Any invented price or policy |
The transcript review is the most valuable twenty minutes of the week. Every conversation the bot fumbles tells you either what's missing from your knowledge sheet or what's unclear on your website, and fixing those improves the site for human readers too. After a couple of months, the questions visitors keep asking are also a ready-made list of pages and tutorials your website should have.
Lead-capture chatbot questions owners ask
Will a chatbot annoy visitors who only want to browse?
Only if it interrupts them. Leave the widget closed by default and show a small prompt on your pricing, contact and product pages after 20 to 30 seconds, rather than a full pop-up on every page. Visitors who want help click it; everyone else carries on. Check your bounce rate on those pages for two weeks before and after launch to see whether the prompt is putting anyone off.
Can the chatbot book appointments or viewings itself?
Often, yes. HubSpot's free chatbot builder can offer a meetings link inside the conversation, and most other chat tools connect to a booking calendar through an integration. Start with the bot collecting the details and offering two or three time slots, then let a person confirm. Move to fully automatic booking once you trust its answers and your calendar is kept accurate.
Do I need a developer to install a website chatbot?
Usually not. Every tool mentioned here gives you either a short code snippet to paste into your site's header settings or a ready-made plugin for WordPress, Wix, Squarespace or Shopify. The work that takes time is the knowledge sheet, the lead questions and the routing. A developer only helps if your site is custom-built and nobody on the team can edit its templates.
What happens to leads the bot collects outside office hours?
They should land in your CRM or inbox straight away with a summary, and the bot should tell the visitor honestly when a person will reply, for example by 9am the next working day. Set a morning check as part of someone's routine. A lead captured at 10pm and answered at 3pm the next day is often a lead lost to a quicker competitor.
Further reads
- Best AI Chatbots for Small Business Websites — A wider comparison of website chatbots if none of these four suits you.
- 9 AI Chatbot Mistakes That Lose Small Businesses Customers — The customer-facing errors that undo a good lead-capture set-up.
- How to Measure Whether Your AI Chatbot Is Actually Working — Deeper reporting once the bot has a few months of conversations.
- How to Reply to Every Enquiry in Under Five Minutes With AI — What happens after capture: getting a fast, useful reply out.
- How to Train an AI Chatbot on Your FAQs, Policies, and Prices — How to structure prices and policies so the bot answers correctly.
- How to Build a Quiz Funnel With AI to Capture Leads — Another way to collect qualified leads when chat isn't enough.
- How Small Tour Operators Use AI for Bookings and Itineraries — How a small tour operator can use AI around its booking system, build private itineraries from a route library, and catch the errors before guests do.
- Can AI Qualify Wedding Leads by Date, Budget and Guest Count? — How AI reads messy wedding enquiries, the five ways dates and budgets get misread, three set-ups by effort, and a back-test on last season's leads.
- AI Client Intake for Law Firms: Qualify Enquiries Out of Hours — Set up overnight AI intake that gathers facts, screens fit and urgency, books consultations and hands a clean summary to a person each morning.
- Instant Roof Estimates on Your Website: Do They Win More Jobs? — How satellite-measured instant roof estimators work, why they raise lead volume but change lead quality, what they cost, and how to test one on your own site.
- The Intake Questions That Make AI Cleaning Quotes Accurate — A grouped checklist of intake questions for cleaning quotes, why each one moves the price, how to verify it, and the pricing rules that stop the AI guessing.
- How Cleaning Companies Use AI to Quote and Book Jobs in Minutes — Instant booking forms, AI triage of messy enquiries and calendar rules: how a cleaning firm gets from enquiry to booked slot in minutes.
- Can a Print Shop Use AI to Price Jobs and Check Artwork? — How a print shop splits the work between AI, its price matrix and preflight software, with prompts, a proofing example and a customer report.
- Best AI Lead Generation Tools for Small Businesses (2026) — Ten lead generation tools sorted by the job they do, with verified 2026 prices, a worked example for each, a costed wholesaler stack and a trial scorecard.
- How to Build a Landing Page With AI That Converts — A brief, a section-by-section prompt and a pre-launch checklist for landing pages that turn visitors into enquiries, with a laboratory's page walked through.
- How to Use AI to Find Out Why Website Visitors Don't Buy — Combine your analytics funnel, free session recordings and exit-survey answers, then have AI rank why visitors leave, with an online bookshop worked through.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: HubSpot free chatbot builder product page and Sales Hub pricing page; Tidio pricing page; Intercom pricing page (Fin outcome pricing); Crisp pricing page; EU AI Act Article 50 transparency provisions. Prices checked September 2026.