An AI agent answers an inbound lead in about 30 seconds, every hour of every day, which is the one thing a human acquisitions team structurally cannot do. That is the whole case for ai speed to lead, and it does not need the statistic everyone quotes at you. The famous claim that responding in five minutes beats thirty by 21 times comes from a 2007 study, is routinely credited to the wrong publication, and the write-ups citing it disagree with each other about how many leads it covered.
You do not need it. Your own numbers answer the question better, and this post is mostly about how to get them.
Nobody selling you an AI dialer has measured your business. The number that matters is your contact rate by response time, and you already have the data to calculate it.
Why the famous statistic will not survive your CFO
The 5-versus-30-minute figures trace to work led by James Oldroyd around 2007, run with InsideSales.com. Secondary write-ups report six companies, 15,000 leads and 100,000 call attempts. Others describe a separate 2011 Harvard Business Review article by Oldroyd, McElheran and Elkington, and the summaries of that one disagree about whether it covered 1.25 million leads at 42 companies or audited 2,241 firms.
Three problems with building a business case on it.
It is nearly twenty years old. Consumer response behavior in 2007 predates the smartphone being universal. A seller filling in a form at 11pm on a phone is a different animal from one at a desktop in 2007.
It is about B2B software leads. Not distressed homeowners. A seller weighing whether to take a cash offer is making a slower, heavier decision than a manager evaluating a SaaS trial.
The attribution is a mess. When a number is misattributed this widely, the person you are quoting it to has probably also seen it debunked.
This is not an argument that speed does not matter. It is an argument that you should stop borrowing someone else's evidence for something you can measure directly in about an hour.
How to calculate what ai speed to lead is worth to you
Four numbers, all of which are already in your CRM.
- 1
Pull every inbound lead from the last 90 days
With the timestamp it arrived and the timestamp of your first outbound attempt.
- 2
Bucket by response time
Under 5 minutes, 5 to 30, 30 to 120, over 2 hours, next day. Five buckets is enough.
- 3
Calculate contact rate per bucket
What share of each bucket you actually reached a human in. Not what share you called, what share answered.
- 4
Calculate appointment rate per bucket
Of the ones you reached, what share booked.
Now you have your own curve. It will not match anyone's published figure and that is fine, because it is the only one that describes your market, your offer and your reps.
The number that falls out is cost per booked appointment by response time. Multiply the lift between your fastest and slowest bucket by your monthly lead volume and you have the annual value of closing that gap. That figure is what justifies spending anything on ai speed to lead, and it is defensible because you produced it.
If your CRM cannot tell you when the first outbound attempt happened, that is the finding. Fix the measurement before buying anything to improve a number you cannot see.
Where the time actually goes
Most teams assume the delay is reps being slow. Usually it is not.
What people assume
Reps are slow to pick up leads
The team needs more discipline
We need to hire another rep
Follow-up is inconsistent
What the timestamps usually show
Leads arrive outside working hours
The notification went to an inbox nobody watches on a Saturday
40% of leads arrive between 6pm and 8am
Attempt one is fast, attempt four never happens
Two of those are a staffing problem. The other two are an availability problem, and an availability problem is the one software solves well.
A rep covering 9am to 6pm on weekdays is available for 45 of the 168 hours in a week. That is 27 percent coverage. Nothing about hiring harder changes the arithmetic, and a seller who filled in a form on Sunday morning has contacted four other buyers by Monday.
What an AI agent does in that gap
The useful version is narrow. It answers, qualifies against criteria you set, books into a real calendar, and hands a human everything it learned.
Answers in seconds
Inbound call, text or form. The reply goes out before the seller has closed the tab.
Asks your qualification set
Condition, timeline, mortgage position, why they are selling. The same questions every time, which is more than most teams manage.
Books into the calendar
Against real availability, with confirmation, so nothing double-books.
Hands off with context
Name, source, what they asked, answers captured, recommended next step. Written for someone with 15 seconds.
What it should not do is negotiate, quote a price, make or imply an offer on a property, or state a value. Those belong to a person, and an agent that strays into them creates a problem no speed improvement pays for. An AI agent built for this job escalates instead.
The handoff is the part that gets skipped
An agent that books a call and hands over a transcript has done half the job. A salesperson with fifteen seconds before dialling needs the four facts that decide how to open, not a wall of text.
The difference between a system people use and one they route around is almost always the handoff format.
What this does not fix
Speed improves contact rate. It does nothing for the three things underneath it.
Lead quality. Answering a bad lead in 30 seconds produces a fast no. Our own audit of a client's paid traffic found an entire placement delivering 182 clicks and zero leads, which no response time repairs. That is a paid ads problem.
Follow-up depth. Attempt one being instant means nothing if attempts two through eight never happen. Most deals in this business are won between the third and the eighth touch, which is a CRM and automation problem.
The offer. If your cash offer is not competitive, speed just gets you rejected sooner.
Speed to lead is a multiplier on a funnel that already works. Applied to one that does not, it multiplies zero.
A sane order to do this in
Measure your current response time by bucket, before changing anything
Fix the notification path so a human knows a lead arrived
Add instant acknowledgment, which costs almost nothing
Point AI at after-hours and weekend coverage first
Extend to business hours only once the after-hours numbers hold up
Re-measure the same five buckets at 90 days
Starting with after-hours is deliberate. It is the window where the comparison is against nothing at all, so the result is unambiguous, and it is the window where your reps were never competing anyway.
Frequently asked questions
What is a good speed to lead time?
Faster than whoever else that seller contacted. There is no universal threshold worth chasing, and the only benchmark that means anything is your own contact rate by response-time bucket, measured over 90 days.
Is the 21x five-minute statistic real?
It comes from a 2007 study led by James Oldroyd with InsideSales.com. It is frequently misattributed to Harvard Business Review, the secondary write-ups disagree about its sample size, and it studied B2B software leads rather than home sellers. Treat it as directional at best.
Does ai speed to lead work for seller leads specifically?
The mechanism does, because a distressed seller contacting several buyers goes with whoever responds and sounds competent. What changes is the script and the qualification set, which should be about condition, timeline and mortgage position rather than budget and team size.
Should the AI call or text first?
Match the channel they used. Someone who filled in a web form at 11pm is usually better served by a text, and someone who called expects a call back. Calling a number that has given you no consent for automated calls is a separate problem covered in our post on
AI cold calling laws.
How fast is too fast?
A reply inside about ten seconds reads as automated even when it is not, which costs trust with a seller who wanted a person. Twenty to forty seconds is quick enough to win and slow enough to feel real.
What if we already answer fast during business hours?
Then your gain is entirely in the other 123 hours of the week, and your measurement should compare weekday-daytime leads against everything else rather than fast against slow overall.
Where to start
Pull the 90 days. Bucket them. The answer is usually that daytime response is fine and everything outside it is a hole, which makes the first move obvious and cheap.
If you want help reading your own numbers, book a free strategy call. We go through where leads arrive, how long they sit, and which gap is costing the most. It is a working session, not a pitch.
*Sources: Lead response research led by James Oldroyd with InsideSales.com, circa 2007, as reported in secondary summaries. Oldroyd, McElheran and Elkington, Harvard Business Review, 2011. Original methodology for both was not verified for this post, which is why neither figure is relied on above. Last reviewed October 2026.*
