# How to handle interested versus curious AI replies
> Learn how to sort LinkedIn replies into interested and curious leads, ask qualifying questions with AI, and hand off the right conversations to a human.
- HTML: https://getkinlead.com/ai-sdr-intent-classification-workflow
- Markdown: https://getkinlead.com/ai-sdr-intent-classification-workflow.md

**Short answer:** To handle interested versus curious replies, define the exact signals that count as buying intent and let AI ask one qualifying question before a human joins. Send people who ask about pricing, a demo or a specific problem straight to you, and keep vague replies in a slower follow-up track until they answer a real qualification question.

## Identify the intent of a reply

Most LinkedIn replies fall into two working buckets: interested and curious. Interested means the person asks for something that moves a deal forward: pricing, a demo, a call, a fit check for their team, or help with a problem they name. Curious means the person acknowledges the message without giving you a reason to spend selling time. They say thanks or ask a broad question with no detail.

Treat every reply as a hot lead and your calendar fills with calls that have no problem and no buyer. Treat every reply as a cold lead and you miss the person who typed 'Can you send pricing?' before lunch.

The distinction has to be written down. If you and your AI use different definitions, the inbox labels become noise. Write one sentence for each bucket and keep it where the AI can read it. For Acme, the interested rule might be: the person asks for pricing, a call, a trial or help with a named problem. The curious rule might be: the person replies with thanks, a general question or no detail about their current process.

**Curious replies**

- Thanks for the add.
- Interesting, tell me more.
- I'm not the right person, try our ops team.
- Cool profile. Let's connect.

**Interested replies**

- Do you have a pricing page?
- How does this work for a 40-person agency?
- Our SDRs spend Friday cleaning a spreadsheet. Can you help?
- I have 15 minutes on Thursday for a demo.

> Never treat 'Tell me more' as a buying signal. It asks you to do work before the other person has told you anything about their situation.

## Set your qualification criteria

Before the AI answers a curious reply, write the minimum facts you need before a human joins. A full budget review can wait. You need enough to know whether the person has the problem you solve and whether a call is useful this quarter.

Use four checks: problem, role, timing and next step. Problem is the pain they name. Role is whether they can influence the purchase. Timing is whether the pain is active now. Next step is whether they will accept a short call or a specific resource. If the reply is vague on all four, it stays curious.

Keep the criteria short enough to repeat. A 12-line rubric the AI can follow beats a 4-page qualification doc it ignores. Write the criteria as questions the AI can ask. Abstract labels give you nothing. 'Are you using a spreadsheet or a CRM today?' gives you a usable answer.

Northwind, a 30-person agency, uses one extra check: the person must mention a client deadline or a hiring plan. That check keeps their founder out of calls with people who are only browsing LinkedIn. Your check will be different, but it should be just as concrete.

| | Criterion | Qualifying signal | Disqualifying signal |
| --- | --- | --- | --- |
| Problem | Names a specific task, cost or delay, for example 'Our SDRs spend Friday cleaning a spreadsheet.' | Says the topic sounds interesting. |
| Role | Can approve a trial or introduce the budget owner. | Says they are researching for someone else with no name or date. |
| Timing | Has a live deadline, launch or quarter-end problem. | Wants to keep an idea for next year. |
| Next step | Accepts a 15-minute call or asks for pricing. | Asks for a brochure with no follow-up question. |

## The AI triage workflow

AI triage is a filter. The seller still makes the call. The AI reads each reply, decides whether it is interested or curious, and either alerts you or asks one qualifying question. The aim is to make the lead do some of the qualification work before you open the thread.

Keep the loop simple. If the reply is interested, stop the AI and notify the human. If the reply is curious, the AI asks one question about the person's current process. If the answer names a real problem or a next step, the lead moves to interested. If the answer stays vague after two attempts, the AI parks the lead and follows up later.

1. **Write the profile in plain language.** Describe the account, the buyer and the problem in 10 lines. Include what a good fit looks like and what a bad fit says. The AI needs examples, not a list of adjectives.
2. **Label replies by intent.** Tell the AI to output interested or curious, plus one sentence of reason. The reason matters more than the label because it shows why the message was sorted that way.
3. **Ask one qualifying question.** For curious replies, ask about the current process, a live deadline or the person who owns the budget. Use a question the person can answer in one line.
4. **Trigger the handoff.** When the reply includes a buying signal, send it to the seller with the label, the reason and the last message. The seller should not have to reread the whole thread.

> One question per message. A curious lead who gets a three-question form will close the chat.

## Handle the curious conversation

When someone says 'Tell me more', do not send a product essay. The person has given you attention. They have not given you context. Your job is to trade a small piece of value for a small piece of information. That keeps the thread moving and gives the AI something to score.

Use a give and get reply: answer in one sentence, then ask one question about their current process. For example, 'We help agencies find buyers on LinkedIn from their own account. Are you doing that search manually today, or using a tool?' The answer tells you whether they have a live process and a reason to care.

- **them:** This looks interesting, tell me more.
- **draft:** I can do that. To make sure I send the most relevant info, are you managing leads manually today or using a tool?
- **them:** We're still doing it manually in spreadsheets, which is a nightmare.
- **draft:** That's why most of our customers start with us. I'll introduce you to Alex, who can show you how to automate that.

```
You are reading a LinkedIn reply for Acme, a B2B lead scoring tool.
Label the reply as interested or curious.
Interested means the person asks for pricing, a demo, a call, a fit check, or names a problem with a deadline.
Curious means the person only says thanks, asks a general question, or gives no detail about their process.
For curious replies, ask one question about their current workflow.
Do not ask more than one question in a message.
Return the label, the reason and the next message.
```

## Review mislabels and improve the rules

The first week will produce mistakes. A person who writes 'Send me a deck' may look curious to the AI and interested to you. A person who writes 'We already use a tool' may look curious and be a dead lead. Review 20 replies by hand each Friday and correct the labels. Then update the prompt or playbook with the exact phrases you saw.

Look for patterns in the mistakes. If the AI keeps calling pricing questions curious, your rule probably says 'asks for a demo' and misses 'asks for pricing'. If the AI keeps calling vague praise interested, your rule is too broad. Write the fix as a new example. Skip the theory paragraph.

- Read the last 20 replies in the curious bucket.
- Mark any reply that named a problem, a timeline or a next step.
- Add the exact phrase to the interested examples.
- Remove any rule that puts the same sentence in both buckets.

> Review the curious bucket. Interested replies get your attention anyway. The mislabels hide in the quiet pile.

## Doing this with Kinlead

Kinlead reads replies from every campaign and puts them in one inbox, classified by intent, with interested replies first. You choose how each campaign handles a reply: hand the conversation to you at the first reply, let the AI reply until there is real interest, or let the AI reply until a meeting is booked with your scheduling link. The AI reply modes send on their own, so use hand-off mode if you want to read every reply before it goes out.

Outreach runs from your own LinkedIn account through Unipile. Each LinkedIn account sends at most one action every 5 minutes, inside the invite and message limits and send window you set. When LinkedIn pushes back on an account, Kinlead pauses invites on that account and emails you.

## Related

- [Sales playbook for LinkedIn outreach](https://getkinlead.com/ai-sales-playbook-for-linkedin-outreach.md)
- [Unified inbox](https://getkinlead.com/features/unified-inbox.md)
- [Outreach safety checklist](https://getkinlead.com/ai-outreach-safety-checklist.md)

## Common questions

**Should I let AI book the meeting or do it myself?**

Let AI book when the offer is repeatable and the qualifying questions are clear. For high-value custom work, use AI to qualify and then book the call yourself so you can hear the context.

**How do I stop AI from sounding robotic during qualification?**

Give it a short playbook in your own voice, with two or three example replies. Tell it to keep messages under two sentences and ask one question at a time.

**What if AI misclassifies an interested lead as curious?**

Review the curious bucket every week. When you find a missed buying signal, add the exact phrase to your interested examples and rerun the prompt on recent replies.

**How many qualifying questions should AI ask before handoff?**

One per message, usually two in total. If the person has not named a problem or a next step after two answers, park the lead and follow up later.

[Create a Kinlead account](https://getkinlead.com/signup) or [book a call](https://getkinlead.com/demo) to talk through pricing and setup.
