One mid-sized business. $5.2 million a year bleeding out. Every dollar of it caused by being slow.

Every enquiry is a race. You just didn't know it was scheduled.

The customer starts it, not you. They filled in your form at 8:40pm, and — here's the part that stings — yours wasn't the only form. They shortlisted three of you. Rang two. 78% of customers buy from the business that responds first. Not the best business. Not the cheapest. The first. Best doesn't get a look-in if best answers two days later.

First in, best dressed. It was true when your grandfather said it. The research just put a number on it.

A new lead is an ember. It's hottest the second it lands, and it starts cooling immediately. The numbers on how fast are brutal — and I'll date and source them, because you know how I feel about vendors who don't.

Harvard Business Review's famous audit clocked the average B2B response time at 42 hours. Not minutes. Hours. Forty-two of them. The MIT/InsideSales research found a lead contacted within 5 minutes is 21x more likely to qualify than one contacted at the 30-minute mark — thirty minutes, not three hours — and Velocify measured a 391% conversion lift just for responding inside the first minute. Against all that: only 7% of companies respond within five minutes at all.

The fine print, honestly: that's largely American research. Australia barely measures this stuff — which tells you something on its own — but physics doesn't change at the border, and the response times we clock in our own audits of Australian businesses would not embarrass the 42-hour average. Three hours here is considered good.

And the phone is worse. 62% of callers won't leave a voicemail. They hang up and dial the next number on the list — and the next number answers.

And here's the modern twist: the phone is only one of your front doors now. Count them.

The lead form from your Google ads. The Meta lead ad that captures details inside Facebook — which someone has to remember to check, or wire up to sync. Messenger. WhatsApp. Instagram DMs. The website chat widget. Email. The phone. That's seven or eight doors — and most businesses have a doorbell on exactly one of them.

The customer doesn't rank these channels. To them, a WhatsApp message is an enquiry, exactly as real as a phone call — often realer, because they chose the channel they actually live in. But inside the business? The phone gets answered eventually. The rest just... sit. A Messenger enquiry from Tuesday. A lead-ad form nobody's synced since the campaign launched. A DM under an ad, asking price, ageing quietly while everyone checks the "real" inbox.

Nobody decided to ignore these people. That's the point — nobody decided anything. Interest is quietly rotting in inboxes no one owns, and every one of those enquiries was paid for at full price.

Now the bit that should make your ad account weep. Think about what your marketing actually did for that 8:40pm enquiry.

Your ad found them. Your ad interrupted their evening, sparked the thought, made the problem feel solvable tonight. You paid Google or Meta real money to warm that person up. And then nobody answered — so they rang the next business on the list, arriving pre-sold. Your advertising did the heavy lifting. Someone else's phone manner closed it.

It gets worse. The business that does answer at 8:41pm doesn't just win by default — they look ten times better than the two that didn't. Answering instantly, after everyone else went silent, reads as competence. As care. As proof. You didn't just pay to generate your competitor's sale.

Run that through your ad maths. Your dashboard says cost per lead: $50. Looks fine. But if slow response means only half your leads ever get a real conversation, your true cost per customer just doubled — and the platforms never show you that number, because the leak happens after the click, off their books and on yours. Cost per lead is what you're charged. Cost per customer is what slow decides.

So what does slow actually cost, all up? Let's stop hand-waving and price it.

We ran a real profile through the audit model we use with clients — a solid, unremarkable mid-sized business: $4,000 average sale, 400 leads a month, $5K a month on marketing, responding in about 3 hours, open 9–5 Monday to Friday, 20 calls a week ringing out, one follow-up attempt per lead. A business you'd drive past and call healthy.

The bleed came back at $5.26 million a year in untapped revenue. The response-time gap alone: $658,000. Leads arriving nights and weekends while the office is dark — and HubSpot's data says 52% of enquiries land outside standard business hours — another $336,000. Unanswered calls putting $4.16 million of annual demand at risk, most of it walking straight to competitors. And $706,000 sitting in the database of already-paid-for leads nobody's re-contacted.

And an uncounted number does something worse than cost you money: it sends you hunting the wrong suspect. You can't miss what you never measure — so when revenue disappoints, you reach for the numbers you can see. Marketing gets told to find more leads. Sales gets told to close harder. Budgets go up, targets go up, the pressure lands squarely on the parts of the business that are already working — all to cover a bleed that was never theirs. The wound sits unwatched between the two of them, in the minutes nobody owns, while the people upstream and downstream of it get squeezed to compensate. That's the true cost of not counting it: not just the lost revenue, but years of fixing the wrong thing.

Here's why this matters more right now than at any point in twenty years — and it connects straight back to what Google just did to your clicks.

AI Overviews are answering searches in place. When that answer box appears, clicks to the top result drop 58%. Impressions up 43%, clicks down — visibility without visitors. Meanwhile the clicks that survive got dearer: average CPCs have gone from $2.32 to $5.42 in a decade.

Put those two lines on the same chart and you get the defining trend of 2026: fewer leads, each worth more. The firehose is becoming a tap. And a tap-fed business has no slack left for slow — when you got 1,000 cheap clicks a month, fumbling half of them was survivable arithmetic. When you get 400 expensive ones, every fumble is a felony against your own P&L. The whole machine — the ads, the content, the funnel you've spent years building — falls over at the last metre if nobody picks up.

Scarce leads make speed the whole game.

Your competitors only have to be faster. Not better. Faster.

Now — the good news, and the reason this piece sits in the middle of everything else I've been writing.

Of all the constraints a business can have, reaction time is the one AI genuinely, provably solves. Not "transformation." Not a pilot. This exact job. Reaction time is a machine problem: machines don't sleep, don't let calls ring out, don't get busy, and don't stop at follow-up number 1.7 when the sale statistically closes after follow-up five.

An agent that answers every enquiry inside 30 seconds — 2am Sunday included. A ringing phone that gets picked up every time, and a missed call that gets a text back in seconds instead of silence. Follow-up that persists politely until there's an answer either way. Every one of those ends in a booked call or a banked dollar, which is exactly the test the last article set: point the AI at the constraint, and the constraint, for most of you, is this.

This is the boring end of AI. It's also the end that pays. The task-manager end is more fun. Fun's expensive.