The first 90 seconds after an enquiry lands

Why do the first minutes matter more than the reply itself?

Because the enquirer is still at their desk. In the minutes after they press submit they're looking at your competitors' websites, and they'll answer whichever phone rings or email lands first. An hour later they're in a meeting. By tomorrow they have booked someone, or the need has passed.

A slow reply doesn't feel like a lost job from the inside. Nobody tells you they went elsewhere. The enquiry just goes quiet, and it gets filed under "tyre-kicker" when it was a buyer who got a faster answer down the road.

The missed call has the same shape. The difference is that a web enquiry sits in an inbox looking safe, so it feels like it can wait.

How do you work out what slow replies are costing?

Use your own numbers. Pull the last three months of enquiries from your inbox and form tool, and for each one note two things: how long until a person sent a real reply, and whether it became paid work.

Then:

• Sort the enquiries into bands by reply time: within 15 minutes, within two hours, same day, next day or later.
• Work out the conversion rate in each band. Paid jobs divided by enquiries.
• Compare the fastest band with the slowest. If there's a gap, multiply it by the number of enquiries in the slow bands and by your average job value.

That's a rough figure, and it's yours, which makes it more useful than any industry average. If your bands show no difference at all, that's worth knowing too. It might mean your buyers aren't comparing, and your time is better spent elsewhere.

Most owners find the bigger surprise is the distribution. They assume they reply within the hour. The inbox usually says something less flattering, especially for anything that landed mid-job or after five.

What has to happen in those first 90 seconds?

Less than you'd think, and none of it needs to be clever. The person on the other end wants to know three things: that you got it, that a real person will deal with it, and when.

A fast first response does this:

1. Acknowledges the enquiry in plain words, using what they actually asked about rather than a generic thank-you.
2. Says who will reply and by when, with a time you'll genuinely meet.
3. Gives them one useful thing now: a booking link, the two questions you'll need answered, or the information they'd otherwise have to wait for.
4. Lands the enquiry somewhere visible to whoever is replying, with the details pulled out, so nobody has to dig through a form email to start.

The fourth step is where most time goes missing. The enquiry arrives, gets read on a phone between jobs, gets marked unread to deal with later, and is found again the next evening.

Which parts can AI take over?

The mechanical parts, which is most of the work.

Reasonable for a system to handle:

• Sending the acknowledgement immediately, written around the actual enquiry.
• Sorting enquiries by type: new work, existing client, supplier, job seeker, spam.
• Pulling out the details into a consistent record: name, contact, location, what they want, any dates.
• Checking for obvious gaps and listing what still needs to be asked.
• Drafting the full reply from your own past replies, prices list and standard answers.
• Telling the right person a draft is waiting, and nudging if it's still waiting an hour later.

Keep with a person:

• Anything that quotes a price, or implies one.
• Any promise about availability or a start date.
• Enquiries that are really complaints, or that come from an existing client who's unhappy.
• Anything sensitive, regulated or outside what you normally do.
• The decision to say no.

That split follows the test in what AI won't do. Drafting a reply happens every day, it fits on one page, and a mistake is cheap to catch before it's sent. Sending an unchecked quote fails the last part of that test.

Where does this go wrong?

In predictable places, all of which are fixable before you switch anything on.

• The acknowledgement promises a time nobody meets. "We'll reply within the hour" is worse than saying nothing if the reply comes tomorrow. Promise what you'll actually do.
• The draft sounds like nobody in particular. If it doesn't read like your business, people notice. Train it on replies you've actually sent.
• Drafts pile up unapproved. A draft waiting four hours for approval is no faster than writing it yourself. Approving has to be a 30-second task on a phone, and someone has to own it.
• The system gets the type wrong. An unhappy client sorted as a new lead gets a cheerful sales reply. Route anything uncertain to a person.
• Forms collect more than they should. Every field is personal information you now hold, and the form needs a collection notice and a privacy policy that covers it. This is general information, so confirm your own privacy obligations.

What to do next

• Pull three months of enquiries and work out your real reply times in the four bands above. That number tells you whether this is worth fixing.
• Rewrite your acknowledgement so it names a person and a time you'll meet.
• Write down your five most common enquiry types and the reply you'd ideally send to each. That's the raw material any draft system needs, and it's useful even if you never automate anything.
• Decide who approves drafts and how fast. Without that, a faster draft changes nothing.

If you'd like a second pair of eyes on where your enquiries stall, the AI Tools Assessment covers this and the tools you already pay for.