AI lead follow-up: answer every lead in minutes, hand the hot ones to a person
By the Codivated team · Updated
An AI lead follow-up agent replies to every new lead within minutes, by text or email, asks the questions a good rep would, and keeps the conversation going until the lead is qualified or clearly not a fit. Qualified leads go to a person with the full conversation. A rep can take over any conversation at any time, and the agent steps back.
Why do leads go cold?
Because follow-up is slow and uneven. Leads arrive at night and on weekends, reps prioritize the ones that look best, and most leads get one or two attempts. An agent doesn't get tired of the tenth follow-up, and it answers at 11pm.
What does the agent actually do?
- Replies fast. Within minutes of a form, call-back request or inquiry, in your tone.
- Qualifies. Asks your questions (timeline, budget, need, location) across a natural back-and-forth, not a form.
- Answers. Uses your approved information to answer common questions, and says so when it doesn't know.
- Hands off. When a lead meets your criteria, or asks for a person, it books the meeting or alerts the rep with a summary.
- Keeps nurturing. Leads that aren't ready get spaced follow-ups instead of being dropped.
- Logs everything in your CRM: messages, answers, status.
What results has this produced?
Our team built lead-nurture agents that worked about 80,000 leads a month over SMS, running in production for 12 months. The agents went back and forth with each lead until it was qualified, then handed it to a person. They qualified about 13% of leads, against about 11% for the system they replaced. They replaced a seven-figure software contract.
Where do people stay in control?
| Moment | Who decides |
|---|---|
| First reply and qualifying questions | The agent, within your approved messages and rules |
| Lead meets your criteria | Hand-off to a rep, with the conversation summary |
| Lead asks for a person, is upset or asks something sensitive | Hand-off right away |
| Rep wants to step in | Rep takes over; the agent stops on that conversation |
| Pricing, discounts, commitments | Always a person |
This is the escalation pattern from our human in the loop guide: the agent works on its own until a rule or its own uncertainty hands the case to a person.
Do you need to follow texting rules?
Yes. Text messaging in the US is regulated (consent, opt-out, quiet hours) and carriers require business messaging registration. The agent should honor opt-outs instantly, only message leads who gave consent, and respect sending hours. Work with your legal team on consent language before launch.
What does it connect to?
Your lead sources (web forms, ad platforms, listing sites), your CRM, a calendar for booking, and an SMS and email provider. Our team has connected agents to Salesforce and custom CRMs.
How is this different from a chatbot on your website?
A website chatbot waits for visitors. A follow-up agent goes after every lead you already paid for, across days and weeks, in the channel they answer. Many teams have both.
How do you start?
Pick one lead source and one qualification goal, write the questions your best rep asks, and run the agent in practice mode, drafting replies a person approves, until the conversations read right. Then let it send on its own within your rules. See also AI sales agents and how to measure AI ROI.
Common questions
Will leads know they are talking to AI?
Tell them. Being upfront builds trust, and some states require disclosure. The agent can introduce itself as an assistant for your team.
Can it handle replies in Spanish?
Yes. It can reply in the lead's language and summarize in English for your reps.
What if it says something wrong?
It answers only from approved information, sensitive topics go to a person, and every conversation is logged and reviewable.
Sources
- Building effective agents, Anthropic.