WhatsApp lead qualification bots: asking without interrogating
A qualification form works because people expect it. The same questions in a chat thread read as an interrogation. What to ask, when, and what to infer instead.
Rebecca Pearson11 min read
On this page
- What we'll cover
- Why form logic fails here
- Ask fewer, better questions
- What to infer rather than ask
- Building it with CodeWords
- Order matters more than content
- Giving before asking
- Routing and what the rep receives
- What qualification cannot do
- Following up on somebody who went quiet
- Making it survive
- Limits worth knowing about
- What to build first
- Frequently asked questions
- Related reading
Qualification questions work on a form because the person filling it in understands the exchange: answer these, get the thing. The same questions asked one at a time in a personal chat thread feel like an interrogation, and people abandon them at a rate that surprises teams porting a working form onto this channel.
The channel does offer something a form cannot, which is that somebody is present and will answer a good question immediately. Using that well means asking fewer questions, in a different order, and inferring most of what a form would have demanded.
What we'll cover
- Why form logic fails here
- Ask fewer, better questions
- What to infer rather than ask
- Building it with CodeWords
- Order matters more than content
- Giving before asking
- Routing and what the rep receives
- What qualification cannot do
- Following up on somebody who went quiet
- Making it survive
- Limits worth knowing about
- What to build first
- Frequently asked questions
Why form logic fails here
A form is a visible contract: the person sees how many fields remain.
A chat has no visible end, so each question raises the question of how many more are coming.
It is a personal channel, where a sequence of demands reads differently from the same sequence in a web form.
People answer and then leave when the exchange feels one-sided.
Abandonment is invisible. They simply stop replying, and a containment metric counts it as a completed conversation.
Porting a seven-field form is the most common and most costly mistake here.
Ask fewer, better questions
Two or three, maximum, before somebody gets something in return.
Ask what changes the response, not what populates a CRM field.
"What are you trying to do" outperforms every structured qualifier and produces information you could not have asked for.
Avoid budget and timeline early, which are answered defensively or not at all on this channel.
One question per message, since two in one message reliably produces one answer.
Accept partial answers and move on, rather than repeating the question, which is the fastest route to abandonment.
What to infer rather than ask
The phone number's country, which frequently establishes region without a question.
Business or consumer from the email domain if you get one, or from how they describe the need.
Urgency from their language — "today", "urgently", a specific date — rather than from a scale question.
The product area from what they described, which is usually explicit in their first message.
Existing customer status from your own records, which should be checked before any question is asked.
Ask only for what you genuinely cannot infer, which is usually one or two things.
Building it with CodeWords
CodeWords connects to more than 3,000 integrations, and the connection is made once and reused.
- Open CodeWords and start a new automation.
- Describe what should happen in plain language to Cody, the automation builder: what qualifies, what is asked, what is inferred, and who is notified.
- Authorize the connection to your WhatsApp Business Platform account and to the CRM.
- Describe the exceptions: an existing customer, an open opportunity, somebody who did not answer, a request for a person.
- Run it against a small volume and read what the conversations actually looked like before scaling.
You describe the outcome; Cody builds it, connects it, and deploys it. The free plan covers light use, with Pro at $39 per month and Business at $100 per month as usage grows; details are on the pricing page.
Order matters more than content
Answer their question first, always, before asking anything.
Establish what you can do for them, so the questions have an evident purpose.
Ask the easiest question first, which builds the pattern of replying.
Leave anything sensitive to last, after the exchange has established some trust.
Ask the hardest question after you have given something, not before.
Stop when you have enough rather than completing a script, which is the discipline that keeps completion rates high.
Giving before asking
Answer the thing they came for, immediately, even if it disqualifies them.
Offer something concrete — availability, a price range, a next step — which justifies the exchange.
Be useful to the unqualified, since they talk to other people and the cost of being helpful is one message.
Do not gate the answer behind the questions, which is the pattern people recognise and resent.
Confirm what you understood, which both checks accuracy and demonstrates listening.
Make the next step obvious rather than leaving them to ask what happens now.
Routing and what the rep receives
Speed dominates on this channel, more than on any other, because the person is present.
Route to an individual and alert them, rather than to a queue.
Send the conversation, not a summary of fields, since the rep can read it and the wording carries information.
Flag what was inferred versus what was stated, so a rep does not repeat something as fact.
Suppress existing customers and open opportunities before anybody is contacted.
Hand over inside the service window where possible, since a reply is free-form and immediate rather than a template.
What qualification cannot do
It cannot make a bad lead good, and disqualifying quickly is a success rather than a failure.
It cannot replace a conversation for anything complex, and attempting it produces a worse version of both.
It cannot assess fit where fit requires judgement about a situation.
It should not score people in a way that determines their treatment without a person involved, which carries obligations in several jurisdictions.
It cannot fix targeting. Qualifying the wrong audience faster produces faster disqualification.
Measure what it actually improved, which is usually response time rather than conversion.
Following up on somebody who went quiet
Most qualification conversations end without a conclusion, and the follow-up is constrained here in a way it is not on email.
The service window closes 24 hours after their last message, after which a free-form reply is not available.
A template is the only route beyond it, which costs money and reads more formally than the conversation did.
One follow-up, not a sequence. The tolerance for chasing on a personal channel is much lower than in an inbox.
Reference the specific thing they asked about, since a generic nudge from a business number reads as marketing.
Make replying trivial — a button, a single question — rather than asking them to re-engage with a process.
Then stop. A second and third template to somebody who did not answer is how a qualification bot damages a quality rating.
Making it survive
Handle the person who stops replying, since that is the most common outcome — record it and leave a route back.
Check the service window before any follow-up, which needs a template outside it.
Suppress against opt-outs and existing relationships on every run.
Keep the conversation attached to the CRM record rather than a summary.
Report the outcome: conversations started, questions answered, qualified, routed, and abandoned by step.
Limits worth knowing about
Abandonment is silent and needs measuring deliberately.
The service window limits follow-up, which changes the design of any multi-day sequence.
Template follow-ups have a cost and a different tone from free-form replies.
People give approximate answers to structured questions on this channel more than on a form.
Automated decisions about people carry obligations in some jurisdictions, including a route to human review.
What to build first
The first exchange: answer what they asked, then ask one thing. It is the shape that determines completion, it is testable within a day, and porting a longer form onto this channel is the error that most reliably produces abandoned conversations nobody counts.
Two habits make the difference. Measure abandonment by step rather than only counting qualified leads, since people simply stop replying and every aggregate treats that as a completed conversation. And answer the customer's question before asking your own, because the exchange has to be visibly two-sided on a channel where the person can see exactly how much they are being asked for.
Frequently asked questions
Why do qualification forms fail when ported to WhatsApp?
A form shows how many fields remain; a chat does not, so every question raises the question of how many more are coming. On a personal channel that reads as an interrogation, and people stop replying rather than objecting.
How many questions can I ask?
Two or three before the person gets something in return, one per message. "What are you trying to do" outperforms structured qualifiers and produces information you would not have thought to ask for.
What should I not ask early?
Budget and timeline, which are answered defensively or not at all on this channel. Leave anything sensitive until after the exchange has established some trust.
What can be inferred instead of asked?
Region from the number, business or consumer from context, urgency from their language, product area from their first message, and existing customer status from your own records — which should be checked before anything is asked.
What order should the conversation take?
Answer their question first, establish what you can do, ask the easiest question first, and stop when you have enough rather than completing a script.
Should the answer be gated behind qualification?
No. It is the pattern people recognise and resent, and being useful to somebody unqualified costs one message and buys goodwill with somebody who talks to other people.
What should a rep receive?
The conversation rather than a summary of fields, with inferred information flagged separately from stated information so nothing gets repeated back as fact.
How do I know whether it is working?
Measure abandonment by step, since people stop replying silently and containment metrics count that as completion. The improvement is usually response time rather than conversion rate.
How do I follow up when somebody stops replying?
Inside the 24-hour window a free-form message is available; beyond it only an approved template, which costs money and reads more formally. One follow-up referencing the specific thing they asked about, then stop.
Why not run a follow-up sequence like email?
Because the tolerance for chasing is far lower on a personal channel. A second and third template to somebody who did not answer is a reliable way to damage the quality rating the whole number depends on.