OptimisedIt is 9.12pm on a Wednesday and someone two streets away has just decided they need exactly what you sell. They have three tabs open, they are working down the list, and you are the one who does not pick up.
I have spent years watching small British firms bleed work in that ten minute window, and the pattern almost never changes. The enquiry does not vanish because your prices are wrong or your reviews are thin. It vanishes because a competitor's system said something back within a minute and yours said nothing at all until half nine the next morning.
The Short Answer, Before I Go Deep
Capturing after-hours leads comes down to two jobs. Something has to respond within seconds so the person stops shopping, and that response has to end with a name, a number, and enough detail that you can pick the job up cold at half seven the next morning. Everything else, the voice agents, the chat widgets, the routing rules, is plumbing built around those two jobs.
You can do this for about £15 a month with a tidy web form and a text autoresponder. You can also do it for £300 a month with an AI receptionist that answers the phone and books straight into your calendar. Both work. Neither works if nobody actually reads the overnight queue before the kettle has boiled.
The rest of this piece covers when UK enquiries really land, what the honest costs are in sterling, what the ICO expects of you, and the specific setup I would build if I were starting on Monday.
When Your Enquiries Actually Arrive
Before you spend a penny, look at your own numbers. In Google Analytics 4, open Reports, then Life cycle, then Engagement, and add hour of day as a secondary dimension across the last 90 days. Do the same with your mobile call log and your inbox. Most owners are genuinely surprised. The peak is rarely Tuesday at two in the afternoon.
British consumer behaviour has shifted decisively towards the evening, and the property market shows it most clearly. Rightmove's own analysis of when the country searches for homes put the busiest hour between 8pm and 9pm, with the single busiest minute of the week at 8.48pm on a Wednesday and the quietest at 4am on a Monday. That data is now several years old and habits have moved further towards mobile since, but the shape holds across almost every consumer-facing sector I have looked at. Dinner, then sofa, then phone, then enquiry.
The phone side is just as leaky. Research by TelePA, covering 142 small enterprises across the UK with a single test call placed to each, found that 47 percent of those calls went unanswered, and that more than a third of the businesses that did answer failed to engage properly with the caller. That study was conducted during ordinary working hours. Outside them, the number heads towards total.
Weekends deserve their own look. If you sell to households rather than offices, Saturday morning between nine and noon is often your second biggest window, and it is one that answering services frequently price as an expensive extra. Bank holidays behave the same way, only more so, and January is brutal for anyone in home improvement because people spend the first fortnight of the year deciding to finally sort the thing that annoyed them over Christmas. Check your enquiry data across a full year before you assume your quiet months are genuinely quiet rather than simply unanswered.
The Statistic Everyone Quotes, And What It Really Says
You will meet the same claim on every vendor site in this market: respond within five minutes and you are 21 times more likely to qualify the lead. It gets attributed to Harvard Business Review roughly nine times out of ten. That attribution is wrong, and it matters, because the people selling you software rarely read the papers they cite.
The 21 times figure, along with the claim that contact odds fall 100 times between a five minute and a thirty minute response, comes from the Lead Response Management study run by Dr James Oldroyd with InsideSales.com in 2007, covering more than 15,000 leads. Harvard Business Review's own contribution was different. In the March 2011 piece on the short life of online sales leads, Oldroyd, Kristina McElheran and David Elkington audited 2,241 US companies and found an average first response time of 42 hours, with 23 percent never responding at all.
Two caveats a British reader deserves. Both studies are American, and both are old. There is no equivalent UK dataset of that scale that I have been able to find, so treat the multipliers as directionally true rather than as physics. The underlying mechanism, though, translates fine: a person comparing three plumbers at 9pm is not conducting a careful procurement exercise. They are ringing down a list until someone responds, and British consumers behave no differently from American ones in that moment.
What I would take from it is simpler and harder to argue with. Speed beats polish. A rough reply in sixty seconds outperforms a beautifully written one at nine the next morning, every single time.
What Your Options Actually Are At Nine In The Evening
There are four realistic routes, and the marketing around them muddles the differences badly.
Voicemail with a decent greeting. Cheap and better than a ring-out, but weak on its own, because most callers hang up rather than leave a message. If you use it, record the greeting yourself, name the business, and tell the caller exactly when you will ring back. Best for: firms with very low overnight volume who want a stopgap while they set something proper up.
Human answering services. A real person picks up, follows your script, takes the job details and emails or texts them through. The quality is genuinely good and the caller always knows they are speaking to a human. It is the most expensive option per call, and 24/7 cover usually sits in a higher tier than the headline price. Best for: solicitors, clinics, and anyone where a caller's first impression carries real weight.
AI voice receptionists. These answer in a couple of seconds, ask your qualifying questions, and increasingly book into a live calendar. They have gone from novelty to normal in about eighteen months in Britain. Best for: trades, garages, salons and clinics with predictable call types and a booking system to write into.
Web chat and forms with automated follow-up. The cheapest and most underrated of the four. A form that fires an instant SMS and email, plus a chat widget that captures a number rather than trying to hold a conversation, catches the entire evening browsing wave. Best for: anyone whose enquiries mostly begin on the website rather than the phone.
One option to strike off your list: Google Business Profile chat. Google removed both chat and call history from Business Profiles on 31 July 2024, and its own support documentation on the change confirms there is no way to message a business through the profile now. I still see UK agencies recommending it. Ignore them.
Real UK Prices, Not Vendor Ranges
Pricing in this market is deliberately fogged, so here are the actual published figures I could verify.
Human answering starts lower than most people expect. Moneypenny's own guide to what call answering costs in the UKputs entry level plans covering a small monthly minute bundle at roughly £25 to £50 a month, mid range plans between £75 and £200, and anything with genuine round the clock cover at £200 upwards. Moneypenny bills per minute, and its smaller schemes up to 150 inclusive minutes carry a three month minimum term, with larger schemes running to six or twelve months. That contract length is the detail buyers miss.
A price comparison published by Sayora and checked on 12 August 2026 lines up several providers on the same page: ReceptionHQ from £15 a month before VAT and usage, Sayora itself at £49 a month flat with unlimited minutes, and AnswerConnect at £205 a month for 150 human receptionist minutes plus a £49.99 setup fee and £1.75 for every extra minute. That same guide cites JAM's buyer guidance that UK answering services generally run between £0.90 and £2 a minute. Sayora sells into this market and appears in its own comparison, which it discloses, so read it as a useful cross-check rather than gospel. You can see the full breakdown of UK answering service costs and follow each provider's link before committing.
AI receptionists in the UK have settled into three bands. Entry products sit around £39 to £80 a month for a small bundle of call minutes. The working mid market is roughly £99 to £299 a month for something that answers, books and syncs with your CRM. Above that you are into bespoke voice agents. Be careful with per-call billing, because a caller who rings back twice about the same job can count as three billable calls. Watch for setup fees and fair usage caps hiding under the word unlimited.
Chat and automation tools are mostly American and bill in US dollars, so every pound figure below is an approximate conversion and will move with the exchange rate. Tidio starts at about $29 a month, roughly £22, with its Lyro AI sold separately from about $39 a month. Intercom charges per seat from about $29 and then $0.99 for every resolution its Fin agent completes, with a 50 resolution monthly minimum, which gets expensive quickly at volume. Zapier's free tier now allows only 100 tasks a month and single step automations, and its Professional plan runs $19.99 a month billed annually, about £15, for 750 tasks. Make bills by operation and works out substantially cheaper at the same volume.
If you want WhatsApp in the mix, Meta has billed per message rather than per conversation since 1 July 2025. UK marketing template messages currently land in the region of four to five pence each and utility messages nearer one and a half to two pence, with replies inside the 24 hour customer window free. That last part changes on 1 October 2026, when Meta begins charging for service messages too, so budget for it.
A Worked Example In Pounds
Vague talk about return on investment helps nobody, so here is the arithmetic I run with clients before they sign anything. Take a domestic electrician in the Midlands, one employee, average job value around £340, and a conversion rate of about one in three on enquiries that actually reach a conversation.
Over a two week count, that firm logged 11 unanswered calls and 6 form submissions that got no reply until the following morning. Call it 17 missed contacts a fortnight, or 34 a month. Not all of those are real work. Roughly a third turn out to be sales calls, wrong numbers, or people who were only ever price shopping. That leaves about 23 genuine enquiries a month arriving when nobody was there.
If a £149 a month AI receptionist captures 80 percent of those, and the firm rings back promptly enough to hold on to a third of what it captures, that is roughly six extra jobs a month at £340, so a little over £2,000 in additional revenue against £149 in cost. The margin on those jobs, not the headline revenue, is what actually matters, but even at 30 percent the system pays for itself four times over.
Now run the same sums for a takeaway with a £22 average order and the answer flips completely. That is the whole point of doing the maths rather than trusting a vendor's ROI calculator, which will always assume you convert everything you capture. You will not.
The Setup I Would Build For A Small UK Firm
If someone handed me a typical five person trades or clinic business tomorrow, this is what I would put in, in this order, over about a fortnight.
Start with the phone, because it is where the money is. Divert your line outside your chosen hours to either an answering service or an AI receptionist, and give it exactly four questions to ask: name, best number, postcode, and what the job is. Nothing else. Long qualifying scripts at 10pm cause hang-ups. Every captured call should arrive as a text to your mobile and an email with a transcript, not as a voicemail file you have to listen to.
Second, fix the form. Cut it to name, phone, and a single free text box. Every extra field costs you completions. On submit, fire an SMS within sixty seconds and an email within two minutes. Both should be plain and human.
Third, add a chat widget only if your site gets meaningful evening traffic, and configure it to collect a phone number early rather than to hold a long conversation. Chat that tries to be clever at midnight generally just annoys people.
Fourth, route everything into one place. A shared inbox works. A free HubSpot or Pipedrive pipeline works better. What matters is that at 7.30am you open one list, not four apps.
Fifth, build the morning routine and defend it. Twenty minutes, first thing, every working day. Ring the overnight enquiries in the order they arrived. This is the step people skip, and skipping it makes the entire spend pointless.
What To Say In That First Automated Reply
Copy matters more than the tooling here, and most defaults are dreadful. The message needs to do three things: confirm a human exists, set a specific time, and give the person somewhere to put more detail.
Something close to this works well: "Thanks for getting in touch with Hartley Plumbing. This is an automated message, but Dave has your enquiry and will call you before 9am tomorrow. If it is an emergency tonight, reply URGENT and we will ring you back within the hour."
Three deliberate choices in there. It says plainly that it is automated, which builds far more trust than pretending otherwise. It names a person and a time, which stops the recipient carrying on down their list. And it offers an urgency escape hatch, because a burst pipe at 11pm is worth waking up for and a quote request is not.
Avoid the word "shortly". Avoid promising a call "within 24 hours", which reads as an admission you are slow. And do not attach a brochure, a discount code, or anything promotional to that first message, for reasons I will come to in a moment.
The Rules That Bite In Britain
This is where imported American advice does real damage, because the UK regime is stricter and the enforcement changed materially this year.
Two sets of rules apply, both policed by the Information Commissioner's Office. UK GDPR governs how you handle the personal data in that enquiry. The Privacy and Electronic Communications Regulations 2003, known as PECR, govern electronic marketing and anything that stores or accesses information on someone's device. The Data (Use and Access) Act 2025 lifted the maximum PECR penalty from £500,000 to £17.5 million or 4 percent of global turnover, and that change took effect in February 2026. Cookie and marketing failures now carry the same financial exposure as a data breach.
The distinction that keeps small firms safe is simple. Replying to an enquiry someone made is a solicited service message, and you do not need marketing consent for it. Adding that person to your newsletter, or slipping an offer into the auto-reply, turns it into direct marketing. The ICO's guidance on electronic mail marketing is unambiguous that the soft opt in exception covers your own previous customers, not prospective ones, and that email and SMS are treated the same way. A new enquirer who has not bought from you is not covered.
Automated outbound calls, meaning recorded voice rather than a live person, need prior consent under PECR. If a vendor offers you an AI agent that rings old leads with a recorded pitch, that is a compliance problem wearing a product badge. Live marketing calls must also be screened against the Telephone Preference Service.
If your system records or transcribes calls, and every AI receptionist does, you need two things. The Telecommunications (Lawful Business Practice) Regulations 2000 require you to make all reasonable efforts to inform callers, which in practice means a short announcement. UK GDPR separately requires a lawful basis, usually legitimate interests supported by a documented assessment, plus a retention period and a privacy notice that actually mentions call recordings. Meeting one duty does not meet the other.
On the website side, the ICO finalised its guidance on storage and access technologies on 29 April 2026, extending the cookie rules explicitly to pixels, scripts, fingerprinting and web storage. Your chat widget almost certainly sets a cookie. It needs to sit behind consent, with accept and reject given equal prominence, and nothing non-essential firing before the visitor chooses.
Where This Falls Over
Honest reality check, because this is not a solved problem and anyone selling it as one is overclaiming.
AI voice agents still stumble on strong regional accents, poor mobile signal, and callers who talk over them. They also handle grief and emergencies badly. If you are a vet, a funeral director, or anything where a 2am caller may be distressed, keep a human path open and test what happens when someone says something the script did not anticipate.
Over-automation is the more common failure. I have watched a firm capture 40 extra enquiries a month and convert fewer of them than before, because every one got a slick automated sequence and none got a phone call. Automation buys you the lead. It does not close it.
The other quiet killer is the unread queue. If your overnight enquiries land in a channel nobody owns, you have paid for a more expensive way of losing work.
Worth flagging one piece of UK infrastructure timing too. Openreach is retiring the old copper telephone network on 31 January 2027, and its own update on the switch off notes roughly 2.8 million lines still to migrate, more than half a million of them serving business premises. If your after-hours plan depends on diverting an analogue line, sort the migration now rather than in the queue next winter.
Questions To Answer Before You Buy Anything
Run through these honestly and the decision usually makes itself.
How many enquiries do you actually miss outside hours? Count a real fortnight. If it is two, buy a £15 answering plan or nothing at all. If it is fifteen, the maths changes completely.
Do they arrive by phone or by form? Phone-heavy businesses need answering cover. Form and chat heavy businesses need speed of reply, and a voice agent is an expensive way to solve a problem you do not have.
What is one job worth to you? A kitchen fitter losing a single £8,000 job pays for two years of most services. A takeaway losing a £22 order does not.
Can the system write into your calendar or CRM, or does it just take a message? Booking capability is the difference between a note and a confirmed job, and it is usually where the price steps up.
Who reads the queue at half seven? If the answer is nobody, fix that before you spend.
What To Measure After A Month
Track four numbers and ignore the vendor dashboard's vanity metrics.
Answer rate outside hours, meaning the share of overnight contacts that got any response at all. Aim for everything.
Median time to first human contact, measured from enquiry to your call, not to the automated text. Under twelve working hours is a reasonable first target for a small firm.
Conversion of after-hours enquiries against daytime ones. If evening leads convert at half the rate, your follow-up is the problem, not your capture.
Cost per captured enquiry. Divide the monthly fee by the enquiries it actually caught. Anything over about a fifth of your average job margin deserves a hard look.
British firms are adopting this kind of tooling fast, for what it is worth. The British Chambers of Commerce found that 54 percent of UK firms are now actively using AI, up from 35 percent in 2025, with around 94 percent of the firms surveyed being SMEs. It also found that 95 percent reported no impact on workforce size, which fits what I see on the ground. Nobody is replacing a receptionist with this. They are covering the fifteen hours a day nobody was covering anyway.
Do This Week
Pick one evening this week and run three test calls to your own number after eight, plus one form submission. Time what happens. Most owners discover something broken they did not know about, usually a divert that never fires or a form notification landing in spam.
Then set a single sixty second SMS autoresponder on your web form. That one change costs almost nothing and typically recovers more after-hours leads than any other single step, because it stops the person carrying on down their list. Add the phone cover next month once you know your real overnight volume rather than your guess at it.