HomeLearnWhen AI answers your leads
What happens when AI answers your leads first? We measured 454 of them.
Most of them hear back in under two minutes, and many book without a person ever getting involved. Across 454 inbound leads in eight of the CRM accounts we run, over the year to September 30, 2026, the typical lead got its first reply in 103 seconds and 68% heard back within five minutes. 62 booked an appointment, and 74% of those booked before anyone from the business had sent a message.
It is not the whole story. Evenings and weekends were slower than business hours, one account lost a third of its text list to opt-outs, and the AI made mistakes a person would not have. All of that is below, with the fix for each.
Four accounts, four very different results.
Same kind of system, different setup. The last column is why.
| Leads | Reached in 5 min | Booked | What made the difference | |
|---|---|---|---|---|
| Roofing contractor | 53 | 79% | 38%, typically 2 minutes after the lead arrived | A text with a booking link the moment the lead lands. 80% of bookings came before any person wrote. |
| Painting contractor | 161 | 79%: 97% in business hours, 53% evenings and weekends | 19%, and 2 in 3 before any person wrote | An AI assistant that answers questions by text and books the estimate. |
| Professional trade school | 111 | 97%, day or night | Not a calendar business: it enrolls students directly | A chat assistant that answers in seconds. The school reports students enrolling without anyone on staff touching the conversation. |
| Our own agency | 100 | 13% | 8% | Mostly webinar registrants, and a 20-minute pause we had built in before our welcome email. We cut it to one minute on September 30, 2026. |
Why so many book without a person.
Speed does most of it. A lead who has just filled in a form is still looking at their phone, and a text that arrives while they are there, with a link to pick a time, gets used. At the roofing contractor the typical booking landed two minutes after the lead did. Across all eight accounts it was thirteen minutes.
That is also why our own number is the uncomfortable one. Our welcome email waited 20 minutes after each registration, so only 13% of our leads heard from us within five minutes. The automation was working exactly as built, and it made us slower than a person would have been. Once this data showed it, we cut the wait to one minute.
Seven things it got wrong, and the fix for each.
It kept calling people who had opted out.
On a review call with the painting contractor, the owner reported calls still going to people who had set Do Not Disturb. The fix: an opt-out now stops calls as well as texts. Email stayed on, because that is what the client wanted.
A reply about rebooking became an opt-out.
A homeowner's message about rebooking appears to have tripped the Do Not Disturb setting by accident. Opt-out words need context, and an ambiguous reply should go to a person instead of silencing a lead who wanted a new time.
Homeowners booked the wrong side of town.
Asked to pick their area on a map, people chose the wrong half of the city, so estimates landed on the wrong crew's calendar. The fix was to ask for the zip code first and route from that.
It called people who had already booked.
Some homeowners booked online before the calling campaign reached them. The calls have to stop the moment a booking exists.
Evenings were slower, and not all of it on purpose.
The painting contractor reached 97% of business-hours leads within five minutes but 53% on evenings and weekends, where some messages appear to have waited for morning. Some of that is quiet-hours policy and some was default settings. Know which of your delays are rules and which are accidents.
Texting hard cost a third of the list.
By the end of the year, 33% of the painting contractor's leads had text messages switched off, and 19 had replied STOP. In the other accounts it was 0% to 6%. Fewer, better texts, and stop at the first reply.
Nobody could say what it earned.
The painting contractor reported about $65,000 in new sales in the campaign's first two months, but tracked sales across three systems, so they could not be tied cleanly to bookings. Decide where a sale gets recorded before the campaign starts.
Check your own in ten minutes.
- Submit your own form on a weekday afternoon and again at 8 PM on a Saturday. Time both replies.
- Open the reply and count the taps to a booked time. More than two is too many.
- Reply "can I move it to next week?" and see whether it books you or opts you out.
- Book, then wait a day. If anything still chases you to book, the stop rule is missing.
- Count your opt-outs as a share of leads with a phone number. Our accounts ran 0% to 6%, and one ran 33%.
Questions people ask
What is an AI appointment setter?
Software that answers a new lead by text, chat or phone, answers their questions, and books them onto your calendar without a person in the loop. In our accounts it usually starts with a text within a minute or two of the lead arriving.
Do leads actually book with an AI?
Many do. In our data 62 of 454 leads booked, 13.7%, and 74% of those bookings happened before anyone from the business had written to them. Among the accounts that book appointments, the rate ran as high as 38% at a roofing contractor. A trade school in the same data enrolls students directly instead of booking them.
What is a good speed to lead?
Under five minutes, every hour of the week. Our accounts reached 72% of business-hours leads within five minutes and 62% on evenings and weekends. The gap is usually settings, not effort.
Does texting leads make them opt out?
Too much of it does. Opt-outs ran 0% to 6% of leads with a phone number in most of our accounts, and 33% in the one that texted hardest. Send fewer messages, make each one useful, and stop when they reply.
How did you measure this?
We read each CRM's contacts, messages, calls and calendar events for the year to September 30, 2026, without changing anything. We counted only real inbound leads, removing imports, staff-entered, internal and test contacts, and measured from the moment the lead arrived to the first outbound message or call.
Founder of ElevAIte CRM. He built the follow-up in these accounts and ran the read of their data in September 2026.
Published · Updated
Keep reading
- ServiceSpeed to leadA real lead, minute by minute, and a test of your own.
- Case studyAI calling and text booking for a painting contractorAbout $65,000 in new sales reported in two months.
- ArticleWhy CRM leads go missing with no errorSilent failures, and the monthly check that catches them.
- ServiceAI phone answeringThe Voice agent, on every call you cannot take.
Find out how fast your leads hear back.
Book a working session and we will test your speed to lead with you, weekday and weekend.
Figures come from a read-only pull of eight CRM accounts on September 30, 2026, covering contacts created October 2025 to September 2026. Three other accounts were left out because part of their follow-up runs in other systems or they belong to a separate business. Sales figures are as reported by the client. Businesses are not named. Results are specific to these accounts and are not a prediction for any other.