Define qualified for one destination
Decide where a lead goes before you decide what to ask. A booking, a call back, a specialist, a nurture list and a polite no each need different details.
Details
Qualified means nothing until you name where the lead goes next. Most sites need five routes. Book: a salon client who wants a Tuesday cut goes straight to the calendar. Call back: a law firm prospect with a custody question needs a lawyer on the phone. Route to a specialist: an HVAC company sends commercial rooftop jobs to a different crew. Nurture: someone pricing a move for next spring gets the basics and a follow-up date. Politely decline: a plumber who does not serve that county says so in one message. Each route needs different minimum details. Pick the route first. Then ask only what that route needs.
Book
Call back
Route to specialist
Nurture
Politely decline
Open Qualification destination map
Qualification destination map
- Inquiry: Capture channel, intent and the minimum identity context.
- Ground answer: Use an approved source and resolve conflicts before responding.
- Attempt action: Apply the permitted workflow with least privilege.
- Verify result: Check the downstream state instead of trusting fluent text.
- Handoff: Send context, attempted actions and the unresolved need to a person.
- Book
- A salon client who wants a Tuesday cut goes straight to the calendar. The chat needs only the service, the time and contact details.
- Call back
- A law firm prospect with a custody question needs a lawyer on the phone, so the chat takes a number and a good time to call.
- Route to specialist
- Some leads belong to another team, like commercial rooftop jobs at an HVAC company, so the chat sends them to that crew.
- Nurture
- Someone pricing a move for next spring gets the basics now and a follow-up date, with no pressure to book today.
- Politely decline
- When you cannot help, such as a plumber asked about a county outside its area, the chat says so in one message and ends.
The question-value test
Test every question before it goes in. Keep it only if the answer changes the route, the business can act on it, the visitor knows it and you need it now.
Details
Put every question through four tests before it goes in the chat. Does the answer change where the lead goes? Can the business act on it? Will the visitor know the answer? Do you need it now, or can it wait for the first call? An auto shop asking for the car's make passes all four, because some shops skip certain brands. Asking for the VIN fails the third test, since few people know it by heart. A question that fails any test comes out. Every extra question is one more place to quit. In a 2024 Federal Reserve survey, small employers named reaching customers and growing sales as their most common operational challenge6.
Does the answer change routing?
Can the business act on it?
Is the visitor likely to know?
Is the data necessary now?
Open Question-value checklist
Question-value checklist
- 1. Does the answer change routing?Keep the question only if a different answer sends the lead somewhere else, such as a commercial job to another crew.
- 2. Can the business act on it?Keep it only if staff will use the answer, like a car's make at a shop that skips certain brands.
- 3. Is the visitor likely to know?Cut questions most people cannot answer on the spot, such as a car's VIN or a property's parcel number.
- 4. Is the data necessary now?If the answer can wait for the first call, leave it out of the chat. Each extra question is one more place to quit.
The table above is the accessible HTML version. Downloads contain the same blank prompts and no hidden scoring.
A progressive qualification flow
Ask one question at a time, accept "not sure" and stop as soon as you know the route. Offer a booking the moment one is possible.
Details
Picture a property manager's website, one question at a time. First, intent: "Are you a tenant, an owner, or looking to rent?" Second, serviceability: "Which building or ZIP code?" If it falls outside the portfolio, say so and stop. Third, urgency: a tenant with water coming through the ceiling gets the emergency line at once, with no more questions. Fourth, scope, only as much as the next step needs. Fifth, contact preference: call or text. Sixth, booking, offered the moment a slot makes sense. Every question accepts "not sure" and moves on. Ask for a phone number only after you know you can help. Stop as soon as the route is known.
Intent
Location or serviceability
Urgency
Scope
Contact preference
Booking
Open Branched flow diagram
Branched flow diagram
- Inquiry: Capture channel, intent and the minimum identity context.
- Ground answer: Use an approved source and resolve conflicts before responding.
- Attempt action: Apply the permitted workflow with least privilege.
- Verify result: Check the downstream state instead of trusting fluent text.
- Handoff: Send context, attempted actions and the unresolved need to a person.
- Intent
- The first question sorts the visitor: 'Are you a tenant, an owner, or looking to rent?' Each answer opens a different path.
- Location or serviceability
- The chat asks for the building or ZIP code. If it falls outside the portfolio, the chat says so and stops.
- Urgency
- A tenant with water coming through the ceiling gets the emergency line at once, with no more questions.
- Scope
- The chat asks only as much about the job as the next step needs, and accepts 'not sure' as an answer.
- Contact preference
- The visitor picks call or text. The chat asks for a phone number only after it knows you can help.
- Booking
- A slot is offered the moment one makes sense, and the chat stops asking once the route is known.
Scoring versus rules
Use plain rules for hard yes-or-no eligibility and a score only to order the leads that pass. Send unknown answers to a person, never to a zero.
Details
Use rules for yes-or-no eligibility and save scoring for ordering. Whether you serve that ZIP code, take that case type or work on that equipment is a hard yes or no. A points total that lets a high score outvote a hard no will book work you cannot do. Once a lead passes the rules, priority signals such as urgency and timing can decide who gets called first. Treat a skipped question as unknown, never as zero, and send unknowns to a person. Keep every rule visible so you can explain why a lead went where it did. Never score on age, name, neighborhood or anything that stands in for a protected class.
Hard eligibility
Priority signals
Unknowns
Human judgment
Open Rule-versus-score table
Rule-versus-score table
| Question | How to check |
|---|---|
| Hard eligibility | List each yes-or-no rule, such as ZIP code or case type, and check that a high score can never outvote a no. |
| Priority signals | Check that urgency and timing only change the order of leads that already passed every rule. |
| Unknowns | Skip a question in a test chat and check that the lead is marked unknown and sent to a person, not scored as zero. |
| Human judgment | Pick a lead and ask the vendor to show why it went where it did. Confirm no rule uses age, name or neighborhood. |
Good enough when, the same for every row above: Record the authoritative result or preserve context for the named human handoff. Do not count generated text as resolution.
Responsible sales behavior
Let the chatbot state only facts you approved, make one relevant offer at most, stay quiet when someone is in distress and accept a no the first time.
Details
Selling well in chat starts with one rule: the chatbot states only facts you approved. Bad: "We can definitely fit you in today." Good: "Let me check the calendar." Make one relevant offer at most. An auto shop booking an oil change can mention a tire rotation it already sells. It should not list five packages. Never pitch someone in distress, such as a tenant locked out at night. Treat "no thanks" as final. When price is unclear, hand it to a person instead of guessing. Disclose that it is a bot. California bans using an online bot to mislead people about its artificial identity to drive a sale, and a clear disclosure is a full defense7. This is not legal advice.
Approved facts only
One relevant offer
Do not pitch distress
Respect rejection
Escalate pricing uncertainty
Open Responsible sales copy examples
Responsible sales copy examples
Data, consent and delivery
Say what you collect, collect the least you need and send it into named fields. Keep typed answers out of analytics and delete old leads on a schedule.
Details
Tell visitors what you collect, inside the chat, before you ask. Then collect the minimum: a name, one contact method and what the route needs. The EU's GDPR principles of data minimization, transparency and limited retention make a sound model even for a US site4. Send each lead into set fields such as name, phone, service and urgency, so nobody retypes a paragraph. Analytics should log events like "chat started" or "route: booking", never the visitor's typed answers. If you have UK visitors, the ICO's cookie rules require consent for non-essential cookies2, and Google's consent mode adjusts tags to that choice3. Block spam on the server. Delete old leads on a schedule. This is not legal advice.
Notice
Minimum collection
CRM or inbox fields
Spam prevention
Retention
Access
Open Analytics event dictionary
Analytics event dictionary
| Question | How to check |
|---|---|
| Notice | Open the chat as a visitor and check that it says what it collects before it asks for anything. |
| Minimum collection | Check that the chat asks for a name, one contact method and only what the route needs. |
| CRM or inbox fields | Submit a test lead and confirm that name, phone, service and urgency land in separate fields, not one paragraph. |
| Spam prevention | Confirm spam blocking runs on the server, then send a junk submission and see that it is stopped. |
| Retention | Ask how old leads are deleted and on what schedule, and confirm the schedule is written down. |
| Access | List who can open lead records and transcripts, and check that analytics log events like 'chat started', never typed answers. |
Good enough when, the same for every row above: Record the authoritative result or preserve context for the named human handoff. Do not count generated text as resolution.
Measure the flow
Measure each step against the one before it. Pair booking numbers with false qualification and response time, because a fast wrong lead is still a wrong lead.
Details
Each step's rate should use the step before it as the denominator, so you can spot the question that loses people. Start rate is chats started divided by visitors who saw the widget. Completion rate is chats that reached a route divided by chats started. Qualified route rate is leads sent to book, call back or a specialist, divided by completed chats. Booking completion is confirmed bookings divided by booking attempts. False qualification is passed leads your team could not serve, divided by all qualified leads. Check it monthly. Then track response time. Across 1.25 million online sales leads at 42 larger U.S. companies, firms that tried to reach an inquiry within an hour were nearly seven times as likely to qualify it5.
Start rate
Completion rate
Qualified route rate
Booking completion
False qualification
Lead response time
First AI Employee fit
First AI Employee answers visitor questions, qualifies inquiries, collects contact details and books on a connected calendar. Confirm the questions, calendar and lead delivery during setup.
Details
Current public capabilities
First AI Employee is a managed website assistant for questions, lead qualification and booking1. It answers visitors' questions about the business, qualifies inquiries and collects contact details. It checks a connected calendar and confirms a booking only when the booking succeeds. If it cannot finish, it collects the details instead. That fits a salon or a moving company that wants the next step to be an appointment, not a form. Google review replies are not available yet.
Implementation questions to verify
Confirm the parts your flow depends on during setup, and keep the answers in writing with a date. Which questions will it ask, and in what order? Which calendar will it connect to, and what happens when no slot fits? Where does a qualified lead land, and in which fields? Who follows up, and how fast? Do not assume a connection to your CRM until the vendor confirms it.
Frequently asked questions
Quick answers on what to ask, in what order and how fast to follow up. Check any vendor's routing or CRM claims against its current public pages.
Details
Frequently asked questions
What is a lead-qualification chatbot?
It is a website chat that collects the least information needed to pick the most useful next step: booking, a callback, a specialist, a follow-up later or a polite no.
What questions should a lead chatbot ask?
Ask only questions whose answers change whether you can help, how urgent it is, where the lead goes or what happens next. Common ones cover what they want, where they are, how soon they need it, how big the job is and how they like to be contacted. The right list depends on your business.
Should a chatbot score leads?
It can, but plain yes-or-no rules for who you serve and where leads go are often easier to explain and check. If you do use a score, write down every factor, leave out sensitive traits and anything that stands in for them, and review the leads it got wrong in both directions.
Can a chatbot book qualified leads?
Yes, when it is connected to a live calendar and set up with your open times and booking rules. It should confirm only after the calendar has saved the appointment, and it should tell the visitor plainly when there is a conflict or a tool fails.
Can First AI Employee qualify website leads?
It qualifies inquiries and collects visitors' contact details. It checks a connected calendar and confirms a booking only when the booking succeeds; if it cannot finish, it collects the details instead. Settle the exact questions, calendar and where leads go during setup.