Building a lead-generation AI agent takes five concrete steps: define 4 to 6 qualifying questions that score intent, connect the agent to a channel such as WhatsApp or web chat, route qualified leads automatically into a CRM, and set time-based follow-up triggers so no lead goes cold. Most builds take 2 to 4 weeks.
How do you build a lead-generation AI agent step by step?
The build has six practical steps, done in order. Skip the qualifying-question step and you get a bot that chats but never sorts a real buyer from a browser. Skip the CRM routing step and your sales team keeps digging through a chat log to find who’s worth calling.
- Map qualification criteria – write 4 to 6 questions that separate a real buyer from someone just browsing.
- Pick the channel – WhatsApp Business API, a web chat widget on your site, or LINE for Japan and Southeast Asia audiences.
- Build the conversation flow – a decision tree or LLM-driven script that asks the questions in order and scores each answer.
- Connect to a CRM – push qualified leads via webhook or native integration into HubSpot, Pipedrive, GoHighLevel, or a custom dashboard.
- Set follow-up triggers – automated nudges for silence, abandoned questions, or leads that qualify but don’t book.
- Run it against real traffic for 1 to 2 weeks, then tune the questions and the qualification threshold.
What qualifying questions should a lead-generation AI agent ask?
A good qualifying flow covers need, timeline, budget, and authority in 4 to 6 questions, because every extra question past that point increases drop-off. Keep answers short – multiple-choice buttons on WhatsApp convert better than open text fields.
- What are you looking to get done? (service or product needed)
- When do you need this – this week, this month, or just researching?
- What’s your budget range?
- Are you the person who makes this decision?
- What’s your location or service area?
- What’s the best way and time to reach you?
Score each answer, for example 0 to 3 points, and set a threshold – anything above roughly 8 out of 12 routes straight to sales, anything below goes into a nurture sequence instead of a live rep’s queue.
How do you connect the agent to WhatsApp or web chat?
Connecting requires a WhatsApp Business Account verified through Meta, plus a JavaScript widget embedded on your website; both channels run the same qualifying logic behind the scenes, so a lead gets an identical experience whether they message you on WhatsApp or type into your site’s chat window.
- Register a WhatsApp Business Account and verify it with Meta – this typically takes 2 to 5 business days.
- Connect through a Meta-approved provider such as Twilio or 360dialog, or through Meta’s own Cloud API.
- Embed a web chat widget, a short JavaScript snippet, so browser visitors get the same flow as WhatsApp users.
- For Japan or Southeast Asia audiences, add LINE using the LINE Messaging API alongside WhatsApp – same questions, different app.
- Pre-approve WhatsApp message templates, since Meta requires templates for any outbound message sent outside a 24-hour reply window.
If phone calls convert better for your business than chat, an English-language AI voice agent can run the same qualifying script over the phone – see Ongito’s AI Voice Agent service. Note that voice agents currently work in English only, while chat agents handle Japanese and other languages in text. Ongito’s AI agents overview shows examples of both deployed for real businesses.
How do you route qualified leads into a CRM?
Route leads by having the agent post structured data – name, phone number, question answers, and qualification score – into your CRM the instant a lead crosses your threshold, and tag the record Hot, Warm, or Cold so reps only work leads worth their time.
| Method | Best for | Setup effort |
|---|---|---|
| Native integration (HubSpot, Pipedrive, GoHighLevel) | Teams already using one of these CRMs | Low, a few hours |
| Webhook plus Zapier or Make | Custom stacks or multiple connected tools | Medium, about a day |
| Custom CRM or dashboard build | Businesses needing tailored fields and pipelines | Higher, project-based |
Proper routing is what turns chat volume into sales pipeline. Ongito built this exact flow for a California real estate business (Yasmine Coccaro), where the qualifying agent plus CRM routing produced a 300% increase in leads and a 400% increase in time on page, because visitors got answered instantly instead of filling out a static contact form.
How do you set follow-up triggers so leads don’t go cold?
Set triggers based on lead behavior, not the calendar – no reply within 60 minutes, a question left unanswered mid-flow, or a qualified lead that didn’t book – and fire each trigger on the same channel the lead was already using.
- No reply within 60 minutes: an automated WhatsApp nudge, something like ‘Still there? Happy to answer anything.’
- Qualified but didn’t book: a calendar link sent automatically after 24 hours, with a second nudge at 72 hours.
- Score below threshold: routed into a monthly nurture sequence instead of the live sales queue.
- After a booked job or sale: an automatic review request keeps the pipeline compounding – see Ongito’s Review & Reputation System.
Speed of the first response matters more than the follow-up sequence that comes after it. Ongito built a WhatsApp lead agent for a Tokyo photographer (Sorede Nacho) that cut inquiry response time from over 15 hours down to under 30 seconds, and that single change was what moved inquiries into bookings.
What does it cost to build and run a lead-generation AI agent?
Cost depends on the build path: DIY no-code tools run roughly $50 to $300 a month in software fees but take weeks of your own time to configure and debug; freelancers typically charge $1,500 to $5,000 per project with little ongoing support; a managed build like Ongito’s Lead Generation service costs $6,000 setup plus $1,400 a month and includes the qualifying flow, CRM routing, and follow-up triggers built in.
| Approach | Upfront cost | Ongoing cost | Time to launch |
|---|---|---|---|
| DIY (no-code builder plus Zapier) | $0 to $500 | $50 to $300/month | 3 to 6 weeks of your time |
| Freelancer | $1,500 to $5,000 | $0 to $200/month | 2 to 4 weeks |
| Ongito managed build | $6,000 setup | $1,400/month | 2 to 3 weeks, CRM and triggers included |
Whichever path you choose, budget separately for CRM software and WhatsApp messaging fees, since those are billed by Meta and your CRM provider, not by whoever builds the agent.
If you want a straight answer on which channel, questions, and CRM setup fit your business, book a free 30-minute automation audit and Ongito will map the build before you spend anything.