Every business phone system in Britain is being switched off on 31 January 2027, and most small business owners are treating it as paperwork. It is not paperwork. It is the only time in twenty years you will get to rebuild how your phone rings without paying for it twice — and roughly nine out of ten businesses are about to spend it buying the same dumb phone system they already had, just over the internet.
BT Business launched its AI Receptionist into this exact moment. bOnline launched one too, bundled with a £19.95-a-month line. Every VoIP provider in the market is running "switch-off ready" campaigns. That tells you something useful: the deadline is real, the buying window is open, and the vendors are all selling the same replacement while disagreeing about what it should do.
This guide is the buying decision, not the migration checklist. What actually replaces your analogue line, what each pattern really costs per month in 2026, the one mistake that turns the deadline into a lost year of leads, and how to size the whole thing in an afternoon.
The Public Switched Telephone Network — the copper pair running from the exchange to your master socket — is retired on 31 January 2027. Openreach has said there is no extension. ISDN2 and ISDN30 go with it. If your business runs on a traditional analogue line, a Featureline setup, an old ISDN block, or a broadband-and-phone bundle from a decade ago, your calls have to move to something carried over the internet.
Three things people get wrong about this:
The switch-off is not asking you to buy phones. It is asking you to decide, once, who answers when you cannot.
Every option on the market in 2026 is one of three patterns. Pick the pattern first; the vendor shortlist falls out of it.
RingCentral, 8x8, Dialpad, Zoom Phone, Microsoft Teams Phone, Ooma, GoTo Connect. You rent a seat per person, calls arrive over the internet, you get an auto-attendant, ring groups, voicemail-to-email, an app on every mobile. Published entry pricing sits at roughly $8–$26 per user per month, with mid tiers at $25–$44. TechRadar's 2026 round-up and independent pricing trackers both land on the same range, which is a good sign the numbers are real.
This is the correct answer for one situation: you employ people whose job is to be on the phone. A dispatcher, an office manager, a front desk. If a human is sitting there anyway, buy seats and give them good tools.
The failure mode: a five-person service business buys four seats it does not need, because the plan looked cheaper than the one with two.
Twilio, Telnyx, Vonage. You rent numbers and minutes and write the call flow yourself. Minutes run about $0.009–$0.015 inbound. This is the cheapest raw plumbing and the most expensive decision, because "you build it" means you own the maintenance, the recordings, the failover, and the person who leaves.
Right for platforms and agencies building a product. Wrong for a plumbing company. Skip it unless you have an engineering function you are already paying for.
Keep a cheap number (or your existing system) and put an AI voice agent in front of it: it answers every call in under a second, 24 hours a day, handles the questions you get asked forty times a week, books the job into your calendar, qualifies or routes, and hands a real human the calls that deserve one. Then it texts you a summary and a transcript.
This is the pattern the market moved to in 2026. BT's version books appointments and writes confirmations into Outlook and Google Calendar. bOnline's AI for Business Calls transcribes, summarises, and scores sentiment for £19.95 a month. CallSphere publishes a flat $49 / $99 / $149 ladder. InTake charges $49.50 a month for salons and takes no commission on bookings. Vapi, Retell and Bland sell the same capability as a platform you configure.
Right for: service businesses where the call is the revenue and nobody is sitting by the phone — contractors, clinics, salons, law firms, property managers, restaurants. Which is most of the small business economy.
Here is what happens at scale when a business "just migrates". They port the number to a cloud system. They configure a nice greeting and a voicemail box. They add an after-hours message that says the office is closed. Total cost: a few hundred pounds of setup, twenty-something per seat, and no change whatsoever in what happens to a caller at 7:12 PM.
And the data on that caller is brutal. Consumer surveys put the share of callers who hang up the moment they hit voicemail at 80–86%, and roughly 82% of those immediately dial a competitor — the figure bOnline built its launch pitch around. Voicemail is not a holding pen. It is a deletion button with a friendly voice.
The other half of the cost is quieter. In the research behind BT's own launch, 58% of small and medium firms spend between 30 minutes and two hours a day answering customer calls, and 85% say those calls interrupt other work. That is not a phone bill. That is the owner doing reception instead of the job they charge for.
Take a five-person plumbing and heating company doing 1,200 inbound calls a month at a $340 average job ticket.
Now price the fix. Three cloud seats at $25 for the people who genuinely need a desk phone is $75. An AI front door on top is roughly $49–$149 flat, not per seat. Total: $124–$224 a month, about $1,500–$2,700 a year. Compare that to a human answering service at roughly $1.10 a minute, where 200 overflow calls at four and a half minutes each is nearly $1,000 a month before anyone has booked anything.
This is why the answer is almost never "pick the cheapest phone system". It is "stop paying humans to do the part a machine does perfectly, and start paying them for the part only a human can do."
Published pricing, pulled from vendor pages and 2026 buyer guides — treat it as the shape of the market, then get an itemised quote for your own numbers.
Every client you already manage has a phone. Nearly all of them are on a legacy line with a deadline they have read about and ignored. The switch-off gives you a legitimate, non-pushy reason to call every single one of them in the same quarter — and the fix is a monthly subscription, not a project fee.
That is why the pattern matters commercially. An agency reselling a cloud phone system competes on price with a thousand IT resellers. An agency reselling a white-label AI receptionist platform is selling the thing the client actually feels: booked jobs at 9 PM and a phone that stopped interrupting the work. Recurring revenue, under your brand, on infrastructure you did not have to build. We wrote the economics up in how to resell AI voice agents for recurring revenue.
A well-set-up front door has a short, boring, verifiable profile: it answers before the second ring; it knows your service area, hours and pricing rules because someone loaded them; it books into a real calendar; it texts the caller a confirmation; it transfers anything unusual to a human in one step; and it emails you a transcript you can read in ten seconds.
If a vendor cannot show you all six on your own call log, you are buying a demo. Our buying guide for AI answering services turns the six into questions, and if you want the ROI maths per vertical there is a full breakdown in AI receptionist cost and ROI for contractors.
No. Number porting is standard practice and usually free. Your existing number moves to the new service on an agreed date — the risk is not losing it, it is porting it too late in a queue. Ask for a written porting date at least six weeks before your switch-off window.
Usually, and by a wide margin, because the pricing models differ. A human answering service bills per minute — roughly $1.10 — so a busy month costs more than a quiet one. An AI front door is typically flat, $49–$149 a month for the whole business. The honest comparison is cost per booked job, not cost per call.
Only if humans need to make and take calls on desk phones or shared numbers. Many small service businesses keep two or three seats for the people who live on the phone and let the AI handle everything else. Solo operators often need no seats at all.
Digital lines depend on power and broadband, so they stop working in an outage. Ofcom requires providers to offer at least one hour of battery backup for customers who depend on the line for emergency calls. Confirm what your provider supplies before the migration date, not after.
Configuration is a day. Porting is two to six weeks. The part that takes planning is deciding what should happen to each type of call — that is an afternoon with your last month of call logs, and it is the only step that decides whether the new system earns anything.
You will replace this phone system once, under a deadline you did not choose. The businesses that come out ahead are not the ones that find the cheapest seat price — they are the ones that use the forced migration to move the front desk from a voicemail box to something that answers, books, and remembers.
That is exactly what we built. Nida puts a white-label AI phone agent, a client portal and a website generator in one platform — so a service business gets an answering front desk on day one, and an agency can resell the whole thing under its own brand. See plans and pricing, or read the 2026 guide to AI voice agents for small business before you talk to a single vendor.
← Back to the blog