Salesforce's State of Sales report finds that sales reps spend only ~28% of their week actually selling, with much of the rest spent qualifying and researching unqualified leads. Pre-qualifying in chat shifts that ratio.
Source: Salesforce, State of Sales
Why qualification belongs in chat, not on a sales call
Every minute a sales rep spends on a not-fit lead is a minute the next fit lead waits. Chatbot qualification fixes that by putting the disqualification step before the human ever sees the lead. The chat asks three questions, scores in real time, routes to the right destination, and only the HOT quadrant lands on a sales rep's calendar.
The 7 steps below are the smallest set that actually works. Skip any and you either over-route to sales (rep frustration) or under-route (missed pipeline).
Define your fit signal
One attribute that cleanly splits your ICP from everyone else. For B2B SaaS this is usually team size or role. For B2C it might be use case or industry. The test: if you only knew this one attribute, could you decide whether to invest sales time? If yes, that is your fit signal. If you need three attributes, your ICP is not sharp enough yet, and qualification will be muddy.
Define your intent signal
One attribute that splits buying-now from buying-someday. The strongest signals: timeline (“rolling out this quarter”), current pain (“our current tool is X and breaking”), or trigger event (“onboarded a new team last month”). Avoid budget as the intent signal in chat. People lie about budget early and tell the truth about timeline.
Write the 3-question script
Question one is fit. Question two is intent. Question three is optional and depends on the answer to one and two. If the lead is high fit and high intent, question three is “What time tomorrow works for a 15-min call?”. If lower, question three captures email and use case. The script should branch, not march straight.
Build the 2x2 routing matrix
Four destinations: HOT (fit + intent) goes to Slack and a calendar link in the chat; WARM (fit, no intent) goes to a 5-email nurture; CURIOUS (intent, no fit) gets a content drip and self-serve link; PASS (neither) gets a thank-you and archive. Routing is dumb on purpose. Every lead lands somewhere within seconds of the chat ending.
Add a fast-path to demo
At any point in the chat, “book a demo” should be a one-tap option. A small share of high-intent leads will refuse to answer qualifiers and just want to talk. Let them. You lose qualification data, you keep the lead. The fast-path button lives in the bottom of the widget on every screen, not buried in question three.
Sync to CRM with the full transcript
Five things must hit your CRM: contact email, fit score, intent score, qualifier answers, and the full chat transcript. The transcript is the most-skipped piece and the most valuable. When sales calls back, “I see you mentioned your team uses Intercom and is rolling out next month” turns a cold call into a continuation.
Review weekly for misfires
Friday review: pull every HOT lead from the week. Did sales agree with the score? If 80% of HOTs were actually HOT, the system is calibrated. If 50% were over-scored, the qualifier questions are too easy or wrong. Tighten one question per week. Within 4 weeks the AI score and the human score should agree 85%+ of the time.
Score in chat is dumb. Score in CRM is rich.
The chat-side score has one job: route fast. It runs on three answers and a 2x2 matrix. That is enough to decide HOT vs WARM vs CURIOUS vs PASS in real time, which is what the conversation needs to do.
The CRM-side score is where you layer enrichment, firmographics, intent data, and historical patterns. The chat score gets sales on the call. The CRM score gets sales the right deal size and forecast accuracy. Mixing them produces a slow, brittle chat that asks 9 questions and converts 30% of where it should.
Frequently asked questions
How many qualifying questions should the chatbot ask?▼
Three, max. One for fit (team size, role, industry), one for intent (timeline, current tool), one optional for budget if it matters in your market. Past three questions, completion rate drops sharply. The trick is each question should branch the next, so the bot feels conversational instead of bureaucratic.
Should I use BANT, MEDDIC, or something simpler?▼
For chatbot qualification, simpler. A 2x2 of fit and intent covers 90% of the routing decision. Fit answers “is this our customer profile?”, intent answers “are they buying now?”. BANT and MEDDIC are sales-call frameworks; trying to compress them into 3 chat questions produces a worse experience and worse data.
What's the right first qualifier?▼
The one that lets you disqualify or branch quickly. For most B2B SaaS that's role or team size. “Are you exploring for yourself or for a team?” instantly splits self-serve vs sales motion. For B2C it's use case. The first question should change everything that comes after it.
How do I keep visitors from dropping off?▼
Three rules: (1) ask one question at a time, never multi-field, (2) use buttons not free text where possible, (3) explain why you're asking when the question gets personal (“Quick one so I can route this to the right person:”). Drop-off after question 3 is normal; drop-off before question 2 means your opener or question one is wrong.
Should the AI score the lead or should my CRM?▼
AI in real time, CRM as the source of truth. The AI needs to score immediately so it can route hot leads to a calendar link in the same conversation. Then the score and answers sync to the CRM where your team can re-score with their own logic. The chat-side score is fast and dumb on purpose; CRM scoring is rich and slow.
What about leads that ask great questions but won't share details?▼
Let them. A small percentage of high-intent leads will not answer qualifiers and will go straight to “can I get a demo?”. Build a fast path: at any point in the flow, “book a demo” should be one tap. You will lose qualification data on those leads, but you will not lose the leads. That trade is correct.
Keep reading
Last updated: May 1, 2026