How to Capture Website Leads After Hours

Many small businesses close at 17:30. Their websites do not. People comparing kitchen fitters, accountants or software tools often do it in the evening, after work, which is exactly when nobody is there to reply. The question gets asked anyway, and what happens next decides whether it becomes a customer or a line in somebody else’s diary.

This guide covers what arrives after hours, why the speed of the first reply matters, and how to set up after-hours lead capture step by step. Then it walks through one invented night and ends with what to measure.

What arrives after you close, and where it goes today

After-hours enquiries tend to fall into a few kinds, and several come from people close to a decision:

• The price question. “Roughly what does this cost?” They are building a shortlist.
• The availability question. “Can you come next week?” They have already decided to buy something.
• The fit question. “Do you cover my area?” or “Does it work with Xero?”
• The quote or call request. They want a person, and they want to know when.

On many small business websites, all four meet the same thing. A contact form sends an email to an inbox nobody opens until morning. The phone goes to voicemail. A chat widget, if there is one, says “We’re offline, leave a message”, which is a contact form with a friendlier face. The visitor learns nothing and leaves, and by the time you reply at 09:15 they may have asked two other businesses the same question. That is what a missed lead after hours usually looks like: not a lost email, but a reply that arrives after the decision has moved on.

Why the first reply decides it

The best-known research on this is from 2011 and still worth reading. In “The Short Life of Online Sales Leads”, published in Harvard Business Review in March 2011, James Oldroyd, Kristina McElheran and David Elkington audited 2,241 U.S. companies by sending each a test enquiry through its website. 37% replied within an hour and 16% within one to 24 hours. 24% took more than 24 hours, and 23% never replied at all. Among companies that replied within 30 days, the average response time was 42 hours.

A separate study by the same team, covering 1.25 million sales leads at 42 U.S. companies, showed what that delay costs. Firms that tried to contact a lead within an hour were nearly seven times as likely to qualify it, meaning a real conversation with a decision maker, as firms that tried even an hour later, and more than 60 times as likely as firms that waited 24 hours or longer. (Figures read in the article on hbr.org on 14 October 2026.)

Treat those numbers as a direction, not a forecast: the research is fifteen years old and your market is not theirs. But it shows where after-hours enquiries sit. A question asked at 21:00 and answered at 09:00 has waited twelve hours, which puts it firmly in the slow group, however good the reply.

What after-hours lead capture has to do

You do not need a night shift. You need something on the website that does five jobs while you are away:

1. 1

   Reply at once. Within seconds, in the same window the visitor wrote in.

2. 2

   Keep the contact details. A name and an email, so the conversation can continue tomorrow.

3. 3

   Answer the question that made them write. From your own prices, areas, lead times and policies.

4. 4

   Offer a next step that does not need you awake. A booked call, a site visit or a consultation, at a time you are really free.

5. 5

   Hand over cleanly in the morning. Whatever needs a person waits in one place, with the whole conversation attached.

An AI chatbot that reads your own content can do the first four and set up the fifth. If your after-hours chats are mostly support questions from existing customers, our guide to after-hours support with an AI chatbot covers that side.

Setting it up, step by step

Step 1: Ask for the details before the chat starts

A short form before the first message turns an anonymous chat into a lead. In EzyConn it is the pre-chat form, and you choose the fields: name, email, phone, department and a consent box. Ask for as little as you will use. A name and an email are enough to follow up; ask for a phone number only if you really will call back, and add a consent box if you want to send marketing later. The email also counts: a chat with an email is a lead created in the reports, which is how you will measure all of this.

Step 2: Give it the answers that decide a purchase

Write down the ten questions people ask before they buy from you. For most service businesses they are a version of: what does it cost, how soon can you start, do you cover my area, what is included, and what happens if something goes wrong. Then make sure each one is answered somewhere the AI can read.

The AI agent answers from your own content. Give it your website address and it finds the linked pages, and you keep the ones you want. Add files such as a price list or brochure (PDF, Word, plain text, Markdown or CSV) and text you write yourself. Where the wording must be exact, usually prices and guarantees, add exact questions and answers, which are used first. Anything that lives behind a login, such as a customer area, goes in as a file or as text.

Prices are the hard one, because many businesses keep them off the website. You do not need an exact price, but you need something: a “from” price, a typical range, or how the price is worked out and what the next step is. A range at 21:00 gives the visitor a reason to book the visit tonight. When the agent does not know, it says so and offers a person. When a teammate then answers a question the AI could not, that reply is captured as a draft answer you approve before it goes live, so the gaps close week by week.

Step 3: Let it book the next step

The strongest after-hours reply ends with something in the diary. From the Starter plan, the agent can offer real appointment times and book them in the chat. Connect a Google Calendar or Microsoft 365 calendar and it reads your busy times, so it only offers slots you are free, and each booking can be written into a calendar you choose. The customer gets a confirmation by email.

Decide what the bookable step is: a site visit or measure-up for a trade, an appointment for a clinic or salon, a 20-minute call for a software company or agency. Set its length and your opening hours. A visitor who leaves with a confirmed time is a much warmer lead than one who leaves with “someone will be in touch”. Our page on booking appointments inside the chat shows how the calendar connection works.

Step 4: Add a service area rule

If you travel to customers, prevent the booking you cannot keep. A service area rule lists the postcodes or postcode prefixes you cover, and the check runs before any time is offered, so a visitor outside your area is told at once. Rules like this, along with checks on contact format and availability, come with workflows on the Starter plan.

Step 5: Send each lead to the tools you already use

The chat is where the lead arrives, not necessarily where you keep it. EzyConn sends events such as a new conversation or a new booking to Zapier, or as a signed webhook to your own endpoint, carrying the name and email from the pre-chat form. From there a Zap can add the person to your CRM, a row to a spreadsheet or a message to a team channel, and the signing secret lets your system check that each delivery came from EzyConn. Follow one test chat all the way through before you rely on it.

Step 6: Plan the morning hand-over

Some questions need you: a discount, an unusual job, a complaint. When a visitor asks for a person at night, the agent offers one, and if nobody is free the visitor sees “All our agents are currently busy. We’ll notify you shortly.” The request reaches your team in the inbox, and in Slack if you use it, with the whole conversation. Routing can assign it by hand, in turn, to the least busy person or to whoever was last active.

Then make reading the overnight conversations the first job of the day, before email. A reply at 08:05 that already knows the visitor’s name, postcode and question is the one that wins the job.

Step 7: Install it and test it as a visitor

The website chat goes live in about 3 minutes: one script tag pasted before the closing body tag, or the step-by-step guide for WordPress or Shopify. Then open your site in a private window one evening and ask it your ten questions. Read the answers as a customer would, and fix the content rather than the chatbot.

Worked example: one night at a kitchen fitter

Hollin Kitchens is a kitchen fitter we made up for this example. Dan runs it with two fitters, covers LS postcodes and closes the office at 17:30. His site has a price page with “from” prices, a page about design visits, and an after-hours chat set up as above. Here are three enquiries from one Tuesday night, each of which used to be a contact form email.

The first enquiry is the one the setup is for: a price question at 21:14 became a booked design visit two minutes later, because the price page gave the AI a “from” price and the calendar gave it real times. The second was stopped by the service area rule before it became a wasted morning, and still reached Dan in case he wanted to make an exception. The third needed Dan, so it waited, but as a named lead with a rival quote attached rather than as a voicemail.

Four mistakes that lose after-hours leads

• A long form before the chat. Six fields at 22:00 on a phone is a reason to leave. Start with a name and an email.
• No prices anywhere. The AI can only answer from what you give it, and a range beats silence.
• A callback with no time. “Someone will call you” is weaker than a booked slot.
• Leaving the gaps open. When your team answers what the AI could not, approve the draft answer that reply creates, and check the topics it does worst on every week.

What to measure

Give it a month before you judge it, and count the same things every week.

Watch the AI’s share and the bookings most closely. If the AI hands over most conversations, the content is missing something, and the topics it does worst on will show where. If leads arrive but few book, look at the step before the booking: usually the price answer. Then check the cost side: what an AI chatbot costs depends less on the headline price than on how the AI is billed, which our guide to AI chatbot pricing works through with real prices.

Where EzyConn fits

EzyConn is an AI agent for your website chat, WhatsApp, email, Telegram and an inbound phone line on your own Twilio number. It answers from your own website, files and written answers, in multiple languages, and offers a person when a conversation needs one. Everything in this guide sits in one place: the pre-chat form, answers from your content, bookings into a connected calendar and the service area rule from Starter, Zapier and signed webhooks, and a morning hand-over in the inbox or Slack with the whole conversation.

The lead generation use case shows the pre-chat form and the buying questions in detail, and how EzyConn works walks through the hand-over. The free plan includes the AI agent and needs no card, and every plan’s limits are on the pricing page. To see what your own site would say at 21:30, start free and ask it your ten questions tonight.

After-hours lead capture: common questions

Keep reading

• After-hours support with an AI chatbot
• AI chatbots for appointment booking
• How an AI chatbot qualifies leads
• How much an AI chatbot costs in 2026
• Lead capture with EzyConn: answer first, then hand over