The finding that matters is not which chatbot you buy. It is when it asks for the email. Bots that ask in the first message convert at roughly the rate of a contact form. Bots that answer three or four questions first, then ask, convert several times better.
That is a scripting decision, not a purchase. Every tool below can do it, and most deployments do not.
The numbers in this market are published almost entirely by companies selling chatbots, and they are flattering. Below are the claims with that caveat attached, the arithmetic to work out what a lead is worth on your site, the real prices, and the five design decisions that separate a bot that collects leads from one that annoys people.
The claims, and who made them
| Claim | Typically stated as | Figure | How to read it |
|---|---|---|---|
| Contact form rate | Visitors who complete a form | 1% to 3% | Broadly consistent across sources, and matches what most sites see. |
| Chatbot rate | Visitors who become a lead | 10% to 15% | Published by chatbot vendors. Treat as a best case on a well-designed flow, not an average. |
| The multiple | Chatbot versus form | 3× to 4× | The most repeated figure in the category, and vendor-sourced throughout. |
| Timing effect | Asking for the email late | 3× to 5× | The one worth acting on, because it is a decision you control rather than a product you buy. |
These are industry figures from chatbot vendor research and roundups, not measurements from Asyntai customers. We include them because you will meet them in a sales conversation, and because the last row points at something useful.
Read the table as a whole and a pattern appears. The impressive multiple is not really "chatbot beats form". It is "a conversation that gives something before it asks for something beats a wall of required fields". A badly designed bot is a form with a cartoon avatar, and it converts like one.
A bot that opens with "enter your email to continue" is a contact form wearing a costume.
What a lead is actually worth to you
Before comparing tools, work out whether this project can pay. Three numbers, one calculation. Here it is on a site doing 10,000 visits a month.
Now run it with numbers that are less kind. At 1,000 visitors, a 3% capture and a 5% close rate, you get 1.5 extra customers a month. If your lifetime value is $200 that is $300, and a $132 per seat tool is marginal.
Do this arithmetic first. It takes ten minutes and it tells you which half of the market you should even be shopping in.
The number people forget: lead quality
Capturing four times as many leads is only good news if your sales team can work them. Four hundred extra leads a month at five minutes each is thirty-three hours of somebody's time.
So the honest version of the calculation has a second line: what share of those extra leads are worth contacting? A bot that captures everybody who says hello produces volume and destroys your team's trust in the pipeline within a month. This is why qualification matters more than capture, and it is the next section.
The five design decisions that decide your result
1. Ask for contact details last, not first
This is the finding at the top of the page and it is worth restating as an instruction. Let the visitor ask something. Answer it. Answer the follow-up. Then ask for an email, with a reason attached: to send the quote, the availability, the spec sheet, the recording.
The reason matters as much as the timing. "Enter your email" is a toll gate. "Where should I send the pricing breakdown?" is a service.
2. Qualify with two questions, not seven
Every qualification framework wants budget, authority, need, timeline, company size, industry and headcount. In a chat window, each question you add loses people.
Pick the two that actually change what you do next. For most businesses that is a version of "what are you trying to do" and "when". Everything else can be found later, by a person, once the lead is real.
3. Route the good ones immediately
A qualified lead has a half-life measured in minutes. If your bot captures a strong lead at 2pm and it appears in a shared inbox that someone checks the next morning, you have built a slower contact form.
Check where leads land and how fast: a notification, a CRM record, a Slack message, a calendar booking in the chat itself. Booking in the conversation is the strongest option available and the least commonly configured.
4. Answer the question that was actually asked
Most visitors who talk to a chat widget want something specific: a price, a delivery date, whether you cover their area, whether you integrate with their system. A bot that deflects those to "let me take your details and someone will be in touch" converts worse than the page it interrupted.
This is where scripted bots fail and where a system that can answer from your real pages and documents changes the outcome. Not because the technology is impressive, but because a visitor who gets a straight answer is far more willing to give you an email.
5. Decide what happens when it does not know
Every bot meets a question outside its knowledge. Three endings are acceptable: hand to a person now, take a message with a promise you keep, or say plainly that it does not know and point at the page that does. The unacceptable ending is a confident wrong answer, which costs you the lead and sometimes the reputation.
What the tools cost
| Tool | Bills by | Published price | Relevant to lead generation |
|---|---|---|---|
| Intercom | Seat, plus AI outcome | $29 / $85 / $132 per seat | Fin AI at $0.99 per outcome. Strong routing and CRM tooling, priced accordingly. |
| Tidio | Conversation | $0 / $24.17 / from $49.17 | 50, 100 and 250+ conversations. Visual flow builder, so scripting a qualification path is straightforward. |
| Crisp | Workspace | $0 / $45 / $95 / $295 | Flat per workspace with AI credits included at each tier. Sales staff cost nothing to add. |
| LiveChat | Agent | $19 / $49 / $79 per person | Human-first. Good if a person qualifies, weak if nobody is at a desk. |
| Olark | Seat | $29 per agent | The AI website agent plan is $400 a month billed annually, first seat included. |
| Asyntai | Message | $0 / $39 / $139 / $449 | 100, 2,500, 15,000 and 50,000 messages a month. Answers from your own pages first, then captures. |
We make Asyntai, so read that row as a disclosure. It is priced per message rather than per seat or per lead, which matters here: a lead generation campaign that works produces more conversations, and that is the meter on several of these products.
The trap: your bill scales with your success
This is specific to lead generation and worth its own arithmetic. Suppose the bot works and conversations triple from 100 to 300 a month.
Outcome pricing is not a bad deal when a lead is worth $200. It is a bad surprise when nobody modelled it. Ask for the rate in writing and multiply it by triple your current volume before you sign.
Answer first, capture second
Asyntai reads your existing pages and documents and answers the visitor's real question, then asks for contact details once it has been useful. That order is the whole finding on this page.
Try it on your websiteWhere a lead generation bot fits, and where it does not
Worth doing
- Considered purchases with a real sales conversation: services, B2B, high-value goods.
- Sites where visitors have specific pre-sale questions you answer the same way every week.
- Traffic arriving outside working hours, which a form captures badly and a person does not capture at all.
- A lifetime value high enough that forty extra customers a month is transformative.
Not worth doing
- Low-value transactional sales where nobody wants a conversation, they want a checkout.
- Very low traffic, where 4% of visitors is two people a month.
- A sales team already behind on the leads it has.
- Sites where the real problem is that the offer is unclear. A bot will collect confusion faster.
Automatic lead qualification, and what it can honestly do
Qualifying leads automatically is the promise that sells these tools. It is real, but narrower than the pitch, and knowing the boundary saves a disappointing quarter.
What automation does well: collecting facts. What are you looking for, how many people, which city, what timescale, which system are you on now. These are questions with short factual answers, and a machine gathers them without getting bored on the four hundredth conversation.
What it does badly: judgement. Whether this person can actually sign, whether the stated budget is real, whether the project will survive contact with their board. Those come from a conversation with a person, and no scoring model reads them off a chat transcript reliably.
So the honest design is: automate the facts, route on the facts, and let a person judge. A bot that tells you a lead is "hot" based on three answers is giving you a number, not a verdict, and treating it as a verdict is how sales teams stop trusting the pipeline.
Three qualification strategies that work
- Route by intent, not score. Rather than a single hotness number, branch on what the visitor said they want. "Book a demo" goes to a calendar. "Just researching" goes to a resource and a nurture sequence. "Existing customer with a problem" goes to support, not sales. This is simple, transparent and it never mislabels anybody.
- Disqualify politely and early. If you do not serve a region, a company size or a use case, say so in the conversation. Everyone's time improves, and a clear no leaves a better impression than a form that goes nowhere.
- Ask the timing question. If you only have room for one qualifier, make it "when are you looking to start". It predicts more about whether a lead is workable than budget or headcount, and people answer it honestly.
Managing the leads after capture
Capture is the part everyone plans and the smallest part of the value. What happens in the next hour decides whether those leads become customers.
Speed of first contact
A lead that filled in a chat at 2pm and hears from you on Thursday has moved on. Whatever your tooling, the first response should be automatic and immediate: a confirmation, a resource, a booking link, something. Then a person follows.
Where the record lands
Decide this before launch, not after the first hundred leads pile up. Three workable patterns:
- Straight into the CRM with the transcript attached, so the salesperson opens the conversation knowing what was said.
- Into a channel your team already watches, with a link to the full record. Fastest to set up, and works well for small teams.
- Booked directly into a calendar, skipping the follow-up entirely. The strongest option where your sales motion allows it.
The pattern that fails is a shared inbox nobody owns. If the lead has no owner within an hour, the automation has moved your bottleneck rather than removed it.
What to do with the ones that are not ready
Most captured leads will not be ready to buy. That is normal and it is where a chatbot beats a form: it already knows what they asked about, so the follow-up can be about that specific thing rather than a generic newsletter.
Keep it simple. Two or three messages over a fortnight, each about the topic they raised, each with an easy way back into a conversation. Then stop. A nurture sequence that runs for six months teaches people to ignore your name.
Five questions to ask before you buy
- Can it answer before it asks? Get a demo on your own website content, not on the vendor's. If it can only follow a script you write, you are buying a form builder.
- Where does a captured lead go, and how fast? Ask to see the CRM record or notification appear live during the demo.
- Can it book a meeting inside the conversation? The strongest possible ending, and not every tool has it.
- What does my bill look like at triple the volume? Get the next tier or the outcome rate in writing.
- What does it do when it does not know? Ask them to show you that path. It is the one nobody demos voluntarily.
What to do this week
- Run the arithmetic. Visitors, current capture rate, close rate, lifetime value. If the ceiling is a few hundred dollars a month, buy the cheapest thing that works or nothing at all.
- Write down your top five pre-sale questions. If a tool cannot answer those from your existing pages on day one, it will not capture more leads than your form does.
- Fix the timing before you fix the tooling. Move the email request to after the answer. That single change is the biggest documented effect in this category and it costs nothing.
- Check your team's capacity. Decide in advance what you will do with four times the leads, because that decision is harder than the purchase.
Sources
Tool prices were read from intercom.com/pricing, tidio.com/pricing, crisp.chat/en/pricing, livechat.com/pricing and olark.com/pricing in September 2026. The conversion figures in the first table come from chatbot vendor research and published 2026 roundups, are not independently verified, and are not measurements from Asyntai customers; we have noted the source type on each row rather than presenting them as findings. Confirm any price before you budget against it.