It answers, in a British voice
A real recorded call runs on the homepage. Sixty eight seconds, unedited, booking a new patient and reading the mobile number back digit by digit.
Check it on the homepageThe opportunity
The phone never stopped ringing.
Somebody stopped answering.
Every other channel a business owns has been automated for twenty years. The website takes orders overnight. The inbox sorts itself. The phone still needs a person sitting next to it, and when that person is with a customer, at lunch, or gone home, the call becomes nothing at all.
A missed call is not a deferred call. The caller rings the next name on the list, and the business never learns it happened. That is the gap England Voice sells into: not a better phone system, but the first one that answers.
Every figure on this page is labelled. Verified is true of England Voice today. Market data is external and sourced. Illustrative model is an assumption, not a result.
What already exists.
Four claims, each checkable on this site in under a minute. There is no customer count here, because there are no customers yet.
It builds itself from a website
A prospect types their address into the demo page. The service reads their site, provisions an agent trained on their own opening hours, services and prices, and hands back a number to ring. No signup, no card.
Check itIt reports back
Every client gets a portal with recordings, transcripts and a summary emailed after each call. Sign-in is a magic link, so there is no password to lose.
Check itIt is priced and published
Ninety nine pounds a month for the entry plan, one hundred and forty nine for the busier one, three hundred to set up, cancel on thirty days. The pricing is on the site, not in a sales deck.
Check itWhy now
Four things became true at once.
None of these alone would be enough. Voice AI that could hold a conversation but cost more than a receptionist would be a demo. Cheap voice AI that stumbled over a postcode would be a liability.
-
Latency stopped being the tell Verified
A caller forgives a machine for a lot, but not for a pause before every sentence. The recording on the homepage answers an interruption mid-word. That is the bar, and it is now clearable on commodity infrastructure.
Demonstrated by the recording published on this site.
-
Employing someone got more expensive Market data
Employer National Insurance rose from 13.8% to 15% on 6 April 2025, and the threshold at which an employer starts paying it fell from £758 a month to £417. Every hire in the country got more expensive on the same day.
HMRC, tax year 2025 to 2026. HM Revenue and Customs
-
The buyers are small and there are a lot of them Market data
1.4 million UK businesses employ someone. Most are far too small for a contact centre and far too busy to answer every call themselves. That is a gap between doing nothing and spending twenty thousand pounds a year.
DBT, business population estimates 2025. Department for Business and Trade
-
Answering by machine stopped being strange Illustrative model
Customers already talk to software to check a delivery or move a booking. The objection England Voice has to clear is no longer "a machine answered", it is "the machine was useless". That is a product problem, not a cultural one.
Judgement, not a measured figure.
Market
Built from the bottom up, not from a market report.
The figures below start from the official UK business count and the prices already published on this site. Nothing here is drawn from a paid industry forecast, because a number you cannot open and check is not evidence.
TAM
Market data
UK private sector businesses
5,700,000 businesses
a year at the £99 entry plan
Every private sector business in the United Kingdom.
An arithmetic ceiling, not a forecast. It answers one question only: how big is the pond. Nobody sells to all of it, and any plan that assumes otherwise is not a plan. Department for Business and Trade
SAM
Market data
Businesses that employ someone
1,400,000 businesses
a year at the £99 entry plan
The 1.4 million UK businesses with at least one employee.
The other 4.3 million are people working for themselves. A business with staff has opening hours, a phone that rings while everyone is busy, and a payroll cost to compare England Voice against. That comparison is the pitch. Department for Business and Trade
SOM
Illustrative model
Appointment-led and callout-led trades
210,000 to 350,000 businesses
a year at the £99 entry plan
Businesses where a phone call is how work actually arrives.
A dental practice, a vet, a garage, a law firm. The call is not admin, it is the booking. Sized here as 15% to 25% of the 1.4 million employing businesses. The 1.4 million is official; the percentage is an assumption and is shown as a range for that reason.
BEACHHEAD
Verified
The eight sectors already built
Eight sectors businesses
Dental, veterinary, legal, estate agency, trades, clinics, salons, garages.
Each already has its own landing page, its own scripted call and its own vocabulary. This is where the product goes first, because the work of specialising for it is already done.
See a sector pageImplied annual values multiply the business count by the published £99 entry plan over twelve months. They describe the size of a market, never England Voice revenue, and no share of any of it has been won.
Who pays
Where a missed call has a price on it.
Four of the eight sectors already built, chosen because each makes a different argument: a patient worth years, an emergency that cannot wait, a single instruction worth more than the subscription, and a tradesman with both hands full.
Dental practices
- Calls involve
- Appointment requests, cancellations, rescheduling, emergency triage, after-hours enquiries.
- An agent can
- Books and moves appointments, takes the caller back, escalates genuine pain to the right person, never gives clinical advice.
- Why it pays
- A registered patient returns for years. Which practice answered first decides who gets them.
Veterinary practices
- Calls involve
- Consultations, vaccinations, repeat prescriptions, out-of-hours emergencies.
- An agent can
- Books consults, takes prescription requests without interrupting the vet, recognises the emergencies that cannot wait and escalates immediately.
- Why it pays
- A frightened owner at nine on a Sunday does not leave a voicemail. They ring the next practice.
Law firms
- Calls involve
- New enquiries, conveyancing questions, case updates, callbacks.
- An agent can
- Qualifies the enquiry, captures the matter type, books the call with the right fee earner, never gives legal advice.
- Why it pays
- One conveyancing instruction is worth more than several years of the subscription.
Trades and home services
- Calls involve
- Callouts, quotes, coverage questions, emergencies.
- An agent can
- Takes the job, captures the address and the fault, tells the caller plainly what the callout covers.
- Why it pays
- The phone rings while both hands are under a sink. A plumber cannot answer and work at the same time.
Potential segments and high-value use cases. England Voice does not claim a customer in any of them.
Business model
A monthly fee, a setup fee, and minutes. Verified
Published pricing, live on the site today.
- Business
- Monthly plan
- AI phone agent
- Calls and minutes
- Recurring revenue
Recurring subscription
The monthly plan is the business. It renews until the client stops it, on thirty days’ notice.
One-off setup
£300 to build and tune the agent. It arrives in month one and offsets the cost of winning the account.
Usage above the plan
Minutes beyond the allowance bill at 45p. Volume growth raises revenue without a new contract.
Tier movement
A client whose volume outgrows Essential moves to Professional. Expansion revenue with no new sale.
Agency channel
Agencies resell under their own name and price. England Voice builds and runs the agents and splits the revenue.
Referral channel
20% of what a referred business pays, for twelve months, published on the partners page.
Prices as published on the pricing page. Extra minutes bill at 45p. Referral terms are on the partners page.
Unit economics
An illustrative model, with the workings shown. Illustrative model
England Voice has no customers yet, so it has no measured CAC, churn or lifetime value. What follows is a model. Move the assumptions and watch the conclusion move with them, which is the only honest way to show numbers a business has not yet earned.
The published Professional price. Real, though the mix across plans is not yet known.
Assumption. The Professional allowance is 800; most accounts will not reach it.
Assumption. Speech recognition, model inference, speech synthesis and telephony combined.
Assumption. Hosting, storage, monitoring and transactional email, per account per month.
Assumption. No retention history exists. 4% implies an average life of 25 months.
Assumption, and the least knowable figure here. Treat it as a target to test, not a result.
Illustrative lifetime value
per account, on the assumptions set here
- Gross profit a month
- Gross margin
- Average life
- LTV to CAC
The arithmetic, in full
-
Direct cost=minutes × cost per minute + other cost -
Gross profit=revenue − direct cost -
Gross margin=gross profit ÷ revenue -
Average life=1 ÷ monthly churn -
LTV=monthly gross profit × average life -
LTV:CAC=LTV ÷ CAC
The same account, multiplied
Accounts
Monthly plan
Recurring a month
Recurring a year
of that is gross profit a year on the assumptions in the model above, before any overhead. Setup fees are one-off and excluded.
Illustrative. England Voice has no customers at any of these counts.
What this model is not
The churn model is a simplification
Dividing one by the monthly churn rate assumes churn stays flat across an account's life, which it never does in practice. Early months churn harder than later ones, so a real cohort curve would give a different answer.
The £300 setup fee is deliberately left out of LTV
Counted properly it offsets the cost of acquiring the account rather than adding to its lifetime value. Excluding it keeps the ratio conservative, and an investor can add it back if they disagree.
There is no overhead in here
This is contribution per account, not company profit. It carries no salary, no engineering time, no premises and no fixed cost of any kind. Gross margin and net margin are very different questions.
The figures are rounded on purpose
A model quoted to the penny is a model pretending to be a measurement. The rounding is a reminder of which one this is.
Only the costs that scale per account are in here
Per minute: speech recognition, model inference, speech synthesis and telephony. Per account: hosting, storage, monitoring and transactional email. The costs that do not scale automatically are the ones that matter most and none of them appear above: onboarding, tuning an agent against real transcripts, support when a call goes wrong, and engineering. Gross margin is a software margin. Net margin has to be earned like any other business.
Defensibility
Nothing here is a moat yet. This is where one could be dug.
The underlying models are available to anybody with a credit card, so the honest answer to "why can't someone copy this" is that today they largely could. What follows is where durability would have to come from, and where the work has started.
StartedVertical depth
Eight sectors, each with its own vocabulary, its own escalation rules and its own idea of an emergency. A vet agent that misses chocolate poisoning is not a general agent with a different greeting.
StartedWorkflow after the call
Answering is the commodity. What happens next is not: the booking written into the diary, the summary sent, the urgent one escalated to a person within the minute.
StartedTime to value
The builder turns a website address into a working agent in about ten seconds, with no signup. Whatever the competitive position, the distance from interest to a phone call that works is short.
PotentialSwitching cost
An agent wired into a diary, a CRM and a routing policy, tuned monthly against real transcripts, is not a thing anybody swaps on a Tuesday afternoon. This grows with tenure and does not exist on day one.
PotentialOperational data
Transcripts show where agents fail and which phrasings book more appointments. Compounding, and constrained: it is client data, handled lawfully, and it is not a licence to do as one likes with it.
StartedDistribution
The agency tier and the 20% referral programme both exist and are published. Channels take years to compound and cost little to start.
PotentialTrust
The published limits page says plainly what the agent will not do. Handing a company’s phone line to software is a trust purchase, and being early to say the unflattering part is worth more than a testimonial.
Alternatives
What a business does instead.
Categories rather than company names. Naming a competitor means verifying what it does today, and a table built from somebody else's marketing page is not research. The comparison a buyer actually makes is with the option they already have.
| What is being compared | Hiring someone | Answering service | Voicemail | England Voice |
|---|---|---|---|---|
| Answers at 3am | No | Usually | Takes a message | Yes |
| Recurring cost | Salary plus 15% NI | Per call or per minute | None | £99 to £149 a month |
| Three callers at once | No | Yes | Yes | Yes |
| Knows your prices | Yes | From a script you write | No | From your own website |
| Books into the diary | Yes | Sometimes, for a fee | No | Yes |
| Sector-specific triage | With training | Rarely | No | Built per sector |
| Time to deploy | Weeks to hire | Days | Minutes | About ten seconds to try |
| Cost of more volume | Another salary | Scales with calls | None | Scales with minutes |
| Reads a distressed caller | Yes | Yes | No | No |
| Greets people at the desk | Yes | No | No | No |
The last two rows are losses, and they stay in the table. A comparison with no losses in it is an advert, and any investor who has read one before will assume the rest of the table is written to the same standard.
Risk
What would sink this.
Written by the person who would lose the most if any of them happened. None of these is hypothetical enough to leave off the page.
The core capability is rented
The speech recognition, the model and the telephony are bought from other companies. Anyone with a card can buy the same parts tomorrow. Everything durable has to be built on top of them, and today most of that is still ahead.
What reduces itReduced by vertical depth and workflow, not by the models themselves.
One bad call costs more than one customer
An agent that mishandles a caller in pain, or gives an answer it had no business giving, damages the client's business and the case for the whole category. The blast radius is larger than the account.
What reduces itHard refusal rules, immediate escalation and a published limits page. Reduced, never eliminated.
A £99 subscription cannot carry an expensive sale
Selling software to small businesses is famously costly relative to what they pay. If the real cost of winning an account lands at the top of the slider in the model above, the economics do not work. Nothing here proves it will not.
What reduces itThe referral and agency channels exist partly because direct sales may not pay for itself.
The incumbents are already inside the building
Practice management software, phone providers and answering services already have the customer, the contract and the data. Any of them can add a voice agent as a feature. Being early matters less than being embedded.
What reduces itArgues for depth in a few sectors rather than breadth across many.
One person
Design, engineering, support, sales and the on-call for a service that answers other businesses' phones, all the same person. That is a ceiling on how fast this can grow and a single point of failure while it does.
What reduces itHonest constraint. It is the first thing funding would change.
Regulated buyers bring real obligations
Health and legal callers hand over information under rules that were not written with voice agents in mind, and their tolerance for error is low.
What reduces itData handling is published. The compliance burden grows with the sectors that pay best.
Next
What turns the assumptions into facts.
Every figure in the model above is a slider because nothing has been measured yet. These are the measurements, in the order they can be taken, and each one replaces a specific assumption on this page.
-
The first paying accounts
Turns published pricing into demonstrated pricing, and finds out whether the £300 setup survives contact with a buyer.
ReplacesRevenue per account
-
A real cost of acquisition
What it actually takes to win one, across the demo builder, the referral programme and the agency channel.
ReplacesCAC, the least knowable figure here
-
Six months of retention
Until an account has had six months to leave, the lifetime value on this page is arithmetic rather than a measurement.
ReplacesMonthly churn
-
Minutes and margin at volume
Real usage against real invoices, which is the only way to know whether the blended cost per minute assumption holds.
ReplacesCost per minute and gross margin
-
One sector proven end to end
Dental is furthest along. The number that matters to a practice is appointments booked, not calls answered.
ReplacesThe beachhead argument
Long term
Answering the phone is the wedge, not the business.
The phone is how a small business finds out what its customers want. Whatever answers it sits on the most valuable conversation the business has all day.
-
AI reception Verified
Answers, books, triages, reports. Live, priced and published.
-
Outbound and follow-up Not built
Reminders, confirmations, the callback nobody got round to making.
-
Qualification and sales Not built
Scoring an enquiry before a fee earner spends an hour on it.
-
Business phone infrastructure Not built
Numbers, routing and policy, not a layer bolted onto somebody else’s system.
-
A voice layer for operations Not built
Every spoken interaction a business has, captured and actionable.
Phase one exists. Phases two to five are direction, not roadmap, and nothing in them is built.
Sources
-
Business population estimates 2025 Department for Business and Trade. Published 2 October 2025.UK, private sector
-
Rates and thresholds for employers 2025 to 2026 HM Revenue and Customs. Tax year 2025 to 2026.UK employers
-
Changes to employer Class 1 National Insurance from 6 April 2025 HM Revenue and Customs. Effective 6 April 2025.UK employers
No paid industry forecast is quoted anywhere on this page. Market sizing is built upward from the official UK business count and the prices published on this site, so every step can be checked.
The pitch is on the site, not in a deck.
Everything claimed here can be checked in a few minutes. Ring the agent on the homepage, build one against your own website, and read the page that lists what it cannot do.
Questions about the model, the assumptions, or anything above: hello@englandvoice.com