Every business has a version of the same graveyard. Leads that came in, got one reply, said something noncommittal like "let me talk to my wife," and then went silent. Nobody decided to abandon them. They fell off the list because the next batch showed up and the day ended.
AI lead follow up exists to fix that specific failure. Not the first response, which gets most of the attention, but everything after it: the second touch, the fourth, the one three weeks later when the person finally has budget. Here is what the research says about how much follow-up a sale takes, why human follow-up decays so predictably, and what separates an AI text message responder that helps from one that annoys people until they opt out.
Follow-up is where deals actually die
The 2007 Lead Response Management study led by Dr. James Oldroyd, then at MIT's Sloan School of Management, with InsideSales.com, analyzed three years of data across six companies, more than 15,000 web-generated leads, and over 100,000 call attempts. It found the odds of contacting a lead drop about 100 times when you call at 30 minutes instead of 5, and the odds of qualifying drop 21 times.
That is the case for speed to lead, and it is airtight. It is also why the unanswered call matters as much as the unanswered form: missed call text back software turns that first dropped call into a live thread instead of a voicemail nobody hears.
What gets lost is the other half of that study. The same research found the odds of contact decrease by over 10 times in the first hour alone, and that after 20 hours, every additional dial a rep makes actually hurts their ability to make contact and qualify the lead. Persistence is not the problem. Persistence applied to a lead you already let go cold is a much worse game than persistence applied early.
So the first touch matters enormously. And for most businesses, the first touch is also the only touch.
The decay pattern in human follow-up
The clearest measurement comes from XANT (formerly InsideSales.com) and its Annual 2014 Lead Response Report, which submitted secret-shopper test leads and recorded every call and email that came back. Of the 9,538 companies that received one:
- —47 percent never responded at all
- —The median number of contact attempts was 1
- —The average was 2.2
- —Only 22 percent responded by phone even once, and of those that did, 36 percent called exactly one time
The most common outcome, after "nothing," was one attempt and done. XANT's earlier analysis across 7,960 companies between 2008 and 2012 put the average at 1.3 call attempts before the rep gave up.
Velocify's data from the seller's side matches. Its sales optimization study "The Ultimate Contact Strategy," built on nearly 3.5 million leads from more than 400 companies, noted that 50 percent of leads are never called a second time, and 59 percent of prospects in the dataset were never sent a single email.
Why does it collapse this way? Not laziness. Follow-up is unrewarded work. The first response has a clear trigger: a lead just came in. The fifth touch, eleven days later, has no trigger at all. It happens only if somebody keeps a list, remembers what was said, and does the unglamorous thing on a Tuesday afternoon when three other leads are ringing.
Lead follow up automation exists to remove that dependence on memory, and lead scoring decides who is still worth the eleventh day. Meanwhile the professional standard climbs. The Bridge Group's biennial sales development benchmark reports tracked average attempts per prospect rising from 5.6 in 2012 to 7.3 in 2014 to 8.2 in 2016. The gap keeps widening.
Drip sequences vs. responsive follow-up
Most businesses that try to solve this reach for a drip. Five pre-written messages, spaced over two weeks, fired on a timer. Better than nothing, and not the same thing as follow-up.
A drip is broadcast on a delay. It does not read. Message three goes out on day eight whether the lead replied on day two, asked a question on day four, or said "we already hired someone." That last case is the one that costs you. Someone who told you they were out and then got two more scheduled messages does not think "clumsy automation." They think "these people do not listen," and they hit STOP.
Responsive follow-up is a conversation with gaps in it. The next message depends on what the last one produced. Somebody who says "call me after the 15th" gets contacted after the 15th, on that subject, not with message four of a template. Somebody who asks about pricing gets an answer, and the follow-up picks up from there.
The distinction shows up in the data. Velocify found that when a prospect was sent more than five emails before contact was made, especially in a short window, the conversion rate was 36 percent lower than for leads reached after one to five messages. Volume without responsiveness does not just fail to help. It works against you.
What good AI follow up does differently
This is where an AI text message responder earns its place, through three capabilities a timer does not have.
Context. It reads the whole thread, not the last message. A lead who wrote "we're getting three quotes, decision is end of month" has handed you the competitive situation and the deadline. The next touch should land before that date and acknowledge the other quotes. A timer knows neither fact.
Timing tied to what was said. "Not until October" is not a dead lead. It is a scheduled one. The follow-up that matters is the one in October referencing the original conversation, and that is exactly the touch human teams never get to, because it lives eight weeks past everyone's working memory.
Knowing when to stop. Persistence has a ceiling, and crossing it is worse than stopping early. A system that follows the conversation can tell the difference between silence and refusal. A drip cannot.
None of this is about being more aggressive. AI lead follow up is not "send more messages." It is "send the message the conversation calls for, on the day it calls for it."
How many touches, and over what window
Here is the most useful number in this body of research. Velocify's study of nearly 3.5 million leads found that 93 percent of all converted leads were reached by the sixth call attempt. Only 7 percent were first contacted on the seventh attempt or later, and those late-contact leads were 45 percent less likely to convert than leads reached in six attempts or fewer.
So the working range is roughly six touches, not the "keep going forever" advice that circulates in sales content. Velocify's recommended email cadence topped out at five messages in the first month, with performance degrading past that.
Timing matters as much as count. Velocify's suggested sequence front-loads hard: first attempt inside the first minute, a second in the 30 to 60 minute range, a third within a couple of hours, and the rest spread out to about day 22. The first minutes carry most of the contact probability. The tail exists to catch people whose timing simply was not that day.
Treat that as a shape, not a rule. A storm-season roofing lead decays faster than a wedding inquiry booked eleven months out. Either way: front-load, taper, cap.
When to stop, and when to hand off to a human
Two stopping rules, and one of them is not optional.
Stop when they ask you to stop. Business texting in the US sits under the TCPA, and text-specific expectations are set out in CTIA's Messaging Principles and Best Practices, updated May 2023, which calls for standard STOP handling and for honoring plain-language requests like "unsubscribe," "cancel," or "please opt me out." The FCC's February 2024 consent revocation order went further, requiring that revocation be accepted in any reasonable manner and processed within a reasonable period not to exceed 10 business days.
Opt-out has to be immediate, obvious, and permanent, and it is your obligation, not your vendor's. Talk to counsel about your program. This is product guidance, not legal advice.
Stop when the sequence has run its course. Past roughly six or seven unanswered touches, you are spending resources on leads that were 45 percent less likely to convert anyway. Park them. A quarterly re-engagement is a more honest motion than a sequence that never terminates.
Hand off to a human before either of those, whenever the conversation earns it: a pricing edge case, an upset customer, a qualified lead asking for a person. The handoff only works if the person taking over gets a summary. A raw transcript to scroll while the lead waits is not a handoff.
Where Kai fits
Kai is FromKai's AI texting agent. On follow-up it does a narrow set of things: responds to inbound texts, follows up on the thread rather than a fixed drip, asks the qualifying questions the business defines, answers from company-approved knowledge, schedules appointments on a real calendar, and hands the conversation to a human with a summary when it should. When someone asks to stop, Kai stops.
If your gap is the booking motion rather than the sequence around it, the AI appointment setter breakdown covers that side, and how Kai books the appointment covers how FromKai runs it. A text comes in. Kai handles it, including on day eleven.
Frequently asked questions
What is AI lead follow up?
AI lead follow up is software that continues the conversation with a lead after the first response: sending the next touch, answering questions that come up, and coming back at the time the lead asked for. Unlike a drip campaign, which sends pre-written messages on a fixed timer regardless of what the lead says, responsive AI follow-up bases each next step on the actual content of the thread.
How many follow-up attempts does it take to reach a lead?
Velocify's sales optimization study of nearly 3.5 million leads found 93 percent of converted leads were reached by the sixth call attempt, and that leads requiring seven or more attempts were 45 percent less likely to convert. Six touches, front-loaded in the first hours and tapering over about three weeks, is a reasonable target for most local-services businesses.
Why do most businesses stop following up after one attempt?
Because follow-up has no trigger. XANT's Annual 2014 Lead Response Report found the median number of contact attempts across 9,538 audited companies was exactly one, with 47 percent never responding at all. Velocify separately noted that 50 percent of leads are never called a second time. The first response is prompted by a lead arriving. The fifth requires someone to remember, days later, with no prompt.
What is the difference between a drip sequence and an AI text message responder?
A drip sends scheduled messages on a timer and does not read replies, so message four goes out even if the lead already hired someone else. An AI text message responder reads the thread, responds to what was actually said, follows up on the timeline the lead described, and stops when the conversation is over. Velocify found conversion rates were 36 percent lower for leads sent more than five messages before contact, so volume without responsiveness carries a real cost.
When should AI follow-up stop and a human take over?
Immediately on any opt-out request, which under CTIA's messaging guidelines includes plain-language phrasings, not just the word STOP. Otherwise, hand off whenever the conversation calls for judgment: a pricing exception, a complaint, or a qualified lead who asks for a person. The handoff should include a summary of the conversation so far so the human is not starting cold.
---
Sources: Lead Response Management study, Dr. James B. Oldroyd (MIT Sloan) with InsideSales.com, 2007. XANT/InsideSales.com, Annual 2014 Lead Response Report (audit of 14,061 companies, 2013 test leads). Velocify, "The Ultimate Contact Strategy" sales optimization study, based on nearly 3.5 million leads from 400+ companies generated in 2012. The Bridge Group, Sales Development Metrics and Compensation Benchmark Reports, 2012, 2014, and 2016 editions. CTIA, Messaging Principles and Best Practices, May 2023. Federal Communications Commission, TCPA consent revocation Report and Order, February 2024.