Every salesperson qualifies. They just do it in their head, on the first call, with two or three questions they have learned to ask early: what do you need, when, and can we actually do it for you? AI lead qualification is those same questions asked by software — on the website at midnight, by text after a missed call, on the phone by a voice agent — with the answers written to the CRM so a human sees a sorted queue instead of a pile.
This article is the design, not the sales pitch: which questions, how to score, how to route, and how to keep it from feeling like an interrogation. It builds on how a website chatbot qualifies and books and on the tag structure from the lead management guide.
What qualification is, and is not
Qualification is deciding how a lead should be handled, not whether it deserves a reply. Everyone gets a reply. Qualification decides who gets a phone call in five minutes, who gets a booking link, who gets a polite pointer elsewhere, and who gets a nurture sequence. Done well, it makes the good leads faster and the wrong ones kinder.
Step 1: Write the three questions
Ask your best closer what they need to know before a call is worth their time. The answers cluster into three dimensions:
- Fit — is this something you do, for this kind of customer, in this place? (“What do you need help with?” / “Which area are you in?”)
- Timing — how soon? (“When are you hoping to get this done?”)
- Scale or budget — a range, framed gently. (“Is this a small job or a bigger project?” / “Roughly what budget did you have in mind — under £X, £X–Y, or more?”)
Three questions. Not seven. Each additional question costs a share of good leads, and everything else can be learned on the call. Make them dropdowns or quick-reply buttons wherever the channel allows; free text is for the message box.
Step 2: Score it simply
Keep the model on one line: points for fit, timing and scale, plus a couple of behavioural signals.
| Signal | Points |
|---|---|
| Service is one you offer, in your area | +3 (0 if not — and route elsewhere) |
| Timing: this month | +3 |
| Timing: next 3 months | +1 |
| Budget in or above your range | +2 |
| Replied to a message or booked | +2 |
| Visited pricing/booking page twice | +1 |
Six or more = hot; three to five = warm; under three = nurture. Store the number in a custom field (“Lead score”) and the band as a tag (hot, warm, nurture). Workflows add and subtract; nobody edits it by hand. This is all a CRM like GoHighLevel needs — no scoring add-on required.
Step 3: Ask on every channel
- Website form: the three questions as dropdowns, plus name and contact. The form feeds the CRM through a webhook (setup here) with each answer in its own field.
- Website chatbot: answers the visitor’s question first, then asks the three, one at a time, with a reason (“so I can point you to the right person”). Offers the booking link to hot leads directly.
- SMS: after a missed call or form, the instant reply can ask one question (“Is this for a home or a business?”); the reply updates the score. Conversation AI can run this exchange if you want it fully automated.
- Voice: an AI receptionist asks the same three on inbound calls and books or transfers based on the answers.
Same questions, same fields, whatever the channel — so the CRM sees one consistent picture of every lead.
Step 4: Route by band
- Hot: ring the owner or salesperson now (call whisper or push notification), task due in 15 minutes, booking link in the instant reply, opportunity created at the top of “New lead”. Automation steps aside.
- Warm: instant reply with the booking link, normal speed-to-lead sequence, task within the hour.
- Nurture: instant reply, a light sequence over three weeks, monthly broadcast afterwards. No call task — the salesperson’s time goes to hot and warm.
- Not a fit: a courteous automated message with the most useful thing you can offer (a referral, a guide), tag
not-a-fit, opportunity closed with the reason. Kind, fast and final.
Step 5: Let humans override
The salesperson who talks to a “nurture” lead and discovers a big project should be able to tag it hot and have the routing respond. The score is a starting point, not a verdict. Keep a “reason” note field so overrides teach you where the questions are wrong.
How it feels to the lead (the part that decides whether it works)
- Help first, then ask. A chatbot that opens with three questions loses people; one that answers “yes, we cover Leeds — same-week appointments” and then asks “is this for a home or a business?” keeps them.
- One question at a time, with a reason.
- Quick replies, not typing. Buttons for the options.
- Never repeat a question the lead already answered on the form. Read the CRM before asking.
- Say what happens next, every time: “Great — Saleem will call you within the hour.”
Measuring it
Two numbers tell you if qualification is working, monthly:
- Booked-call rate by band. Hot should convert several times better than nurture. If it does not, the questions or the scoring are wrong.
- Salesperson time per won deal. Should fall, because their calls are now with the right people.
And one guardrail: the share of leads marked “not a fit” that later became customers elsewhere in your data. If that number is more than a handful, your fit question is too strict.
Mistakes I see
- Seven questions. Three, then the call.
- Scoring nobody uses — if routing does not change by band, the score is decoration.
- Different questions per channel, so the CRM has three incomplete pictures.
- Hot leads sitting in a queue because the notification went to “everyone”.
- No human override, so the model’s mistakes become policy.
Start with the form. Three dropdowns on the website form, a score field, and one routing rule (hot → ring now). You can add the chatbot and voice layers once you have seen the bands work for a month.
Further reading: GoHighLevel Help Center.
FAQ
What questions should an AI ask to qualify a lead?
The two or three your best salesperson asks first: what they need, when they need it, and whether they are in your area or budget range. Anything beyond that belongs on the call, not in the chatbot.
Will qualification questions put leads off?
Only if they come before help. Answer the visitor’s question first, then ask one thing at a time with a reason (“so I can point you to the right person”). Framed that way, most people answer.
Do I need lead scoring software?
No. A CRM with tags and custom fields — GoHighLevel, HubSpot, most others — can hold a score that workflows update. Dedicated scoring tools only make sense at large lead volumes.
Should hot leads skip the automation entirely?
Yes. A lead that answers “need it this week, in your area, budget confirmed” should ring a human’s phone immediately. Automation should qualify, then get out of the way.
Want your three questions, scoring and routing designed and built into your CRM and website? Book a call — bring your best salesperson’s first-call questions; that is the whole spec.
