If you run a small business, your phone is probably your most expensive unanswered employee. On October 1, 2026, BT Business — one of the United Kingdom’s biggest telecom companies — launched an AI Receptionist for small businesses, and that matters even if you are in the United States and cannot buy it. This guide explains what BT actually launched, what the best AI receptionist for small business looks like on the US market today, what it costs, and the trust rules to set before an AI voice answers a real customer.
Table of Contents
What BT actually launched on October 1
BT’s AI Receptionist is the first product in its new “AI from BT” portfolio, launched alongside a free AI Adoption Hub with practical guides and examples. According to BT’s announcement, the service answers calls around the clock, handles routine enquiries, schedules appointments and sends confirmations, and integrates with Outlook and Google Calendar. It captures and transcribes conversations, draws insights from them, and offers customizable voices and accents, learning from the business’s own information.
Setup is through a self-service portal, on flexible contracts, and BT says the service works across any UK fixed or mobile network, so a business keeps its existing number and phone system. BT also states that customer data is not used to train AI models — a promise worth demanding from any vendor.
Two hard limits for US readers: BT has not announced pricing, and the service is UK-only — you cannot buy it from the United States. Its value here is as a signal and a checklist: demand the same features — 24/7 answering, real booking, transcripts, calendar integration, and a no-training-data policy — from the US vendors below.
The missed-call math: run your own numbers first
BT’s case rests on a survey, and you should read it as a survey, not a measurement. BT commissioned Censuswide to survey 1,500 UK business decision makers and 2,000 UK adults in late September 2026. From those answers, BT estimates UK small businesses lose £3.67 billion a year to missed calls, and says 67% miss at least one potentially valuable call on an average day. That figure is businesses’ own estimates scaled up to a full year — directional evidence, not a measured total of lost revenue.
US data tells a similar story with better documentation. Upfirst’s State of Voice AI for Small Businesses 2026, published October 14, 2026, is based on surveys of 270 US consumers and 262 US business owners and managers, plus 677,562 calls handled for 2,967 US businesses between January 2025 and August 2026. It found that 78% of surveyed US consumers had hired or purchased from a competitor because a business failed to answer or return their call. Among businesses with 10 to 49 employees, 91% missed at least one potential-customer call in a typical week, and 76% missed four or more.
Do your own version of this math before you buy anything. Count last month’s inbound calls, then how many went to voicemail or went unanswered. Multiply the missed calls by your average job value and a conservative booking rate — say, one in five. If you take 100 calls a month, miss 30, and one in five would have become a $200 job, that is $1,200 a month walking away. That example is illustrative, not a promise. If your missed-call count is near zero, you do not have this problem, and an AI receptionist is the wrong purchase.
What US small businesses can actually buy
US vendors meter services in three ways, and the metering matters more than the headline price. Per-minute plans charge for talk time, per-call plans for each conversation, and per-caller plans, like Goodcall’s, for each unique customer in a month — so a repeat caller who phones three times still counts once. Long calls punish per-minute plans; many repeat callers favour per-caller plans; short, simple calls may suit per-call plans.
The table uses September 2026 pricing from an independent comparison verified against vendor pricing pages. Prices change often, so confirm on the vendor’s own page before you commit.
| Vendor | Best for | Entry price | Metering |
|---|---|---|---|
| Rosie | Solo owners and trades | $49/month (250 minutes); $149/month books appointments | Per minute |
| Goodcall | Repeat-caller local services | $79/month (100 unique callers) | Per unique caller |
| Smith.ai AI | Professional services needing human backup | Free (25 calls/month); Pro $150/month (75 calls) | Per call |
| My AI Front Desk | Appointment-heavy practices | $99/month | Per plan tier, scheduling-first |
Rosie suits a solo operator or trade business. Goodcall suits restaurants, retailers, and service businesses where the same customers call again and again, because unlimited minutes under a per-caller cap removes the fear of a long call. Smith.ai is the pick when one mishandled call could cost thousands — a law firm or consultancy — because live agents back up the AI, though its human plans start at $300 a month. My AI Front Desk fits appointment-driven practices where scheduling, not message-taking, is the whole point.
Watch the entry tier: Rosie’s $49 tier takes messages, while real calendar booking starts at $149. Price the tier that does the job you need. Our earlier guide to how an AI phone answering service works explains why message-taking is a different product from booking jobs.
How to choose in 15 minutes
Answer these five questions before any free trial, and you will eliminate most of the market in one pass. First, how many calls and how long are they? Last month’s numbers decide which metering model is cheapest. Second, must it book, or is message-taking enough? If booking matters, test it on your real calendar, not a demo calendar. Third, what happens when the AI does not know? The right answer is a clean transfer to a person or a callback promise with a time attached — never a confident guess. Fourth, does it connect to the calendar and customer system you already use? A transcript in a separate dashboard is another inbox to check. Fifth, what does it do with your call data? Ask for the no-training-data promise in writing, and check how long transcripts are kept.
Run the trial like you would judge an employee. Start with overflow routing only — the AI answers when your team cannot. Make test calls at night, on a weekend, and with a difficult question, and judge it on booked appointments and cleanly captured messages, not on how many calls it answered.
Trust rules before it answers a real call
This is the section most buyers skip, and it protects your reputation. Upfirst’s survey found that 80% of US consumers want an AI receptionist to identify itself as artificial intelligence at the start of a conversation. Do that, in plain words, on every call. Comfort drops fast as tasks get sensitive: 70% were comfortable with hours, services, and pricing questions, and 62% with taking a message, but only 45% with booking or changing appointments, 16% with taking payments, and 15% with handling a complaint. Set permissions to match: hours, pricing, messages, and simple bookings are fair game; payments and complaints go straight to a person.
As BBC News reported on November 2, 2026, its Business Daily programme asked whether businesses can trust AI output, looking at why AI systems sometimes produce confident but false information. Apply that question to your phone line: an AI receptionist’s output is a draft, never a deliverable. Review transcripts and bookings daily for the first month, and give staff a written rule for which tasks the AI may do unsupervised and which it may not. A UK GP surgery’s AI receptionist drew mixed patient reactions when reported by the BBC in October 2026 — a reminder that even a working system can feel wrong if nobody explains it or checks it.
Two more rules belong in writing before launch. Never let the AI quote a price, promise a deadline, or confirm availability it has not verified against your real calendar or price list. And keep a human escape route on every call: a caller who says “person,” “manager,” or “this is urgent” should reach a person or a guaranteed callback, not another menu. For a fuller framework, see our AI for small business guide.
Verdict: who should buy, and who should skip
Buy an AI receptionist if you regularly miss calls during work hours, after hours, or while your team is on site, and if most of those calls are routine: bookings, hours, pricing, and messages. Start with Rosie or Goodcall for simple, high-volume phone work, and Smith.ai when a single mishandled call is expensive. Skip it if you answer nearly every call already, if your calls are mostly complex or emotional — complaints, legal advice, medical triage — or if nobody will review transcripts and bookings daily. Then the AI does not save time; it moves work from answering the phone to cleaning up after it.
BT’s launch does not give US owners a new product to buy. It gives them proof that AI receptionists are now mainstream infrastructure and a feature checklist to hold US vendors to. The businesses that benefit first will be the ones that measured their missed calls, picked the metering model that fits, and set trust rules before the AI ever picked up.
Is BT’s AI Receptionist available in the United States?
No. It launched October 1, 2026 for UK small businesses on UK fixed and mobile networks, and BT has not announced US availability or pricing. US owners should use vendors such as Rosie, Goodcall, or Smith.ai, and use BT’s feature list as a checklist.
How much does the best AI receptionist for small business cost?
Based on September 2026 vendor pricing, entry plans start at $49 a month for Rosie (250 minutes), $79 a month for Goodcall (100 unique callers), and free for 25 calls a month on Smith.ai’s AI plan, with paid plans from $150 a month.
Will customers know they are talking to an AI?
They should. In Upfirst’s 2026 survey, 80% of US consumers wanted an AI receptionist to identify itself at the start of a conversation. Tell callers plainly, keep a human transfer route open, and do not let the AI handle payments or complaints unsupervised.
What should an AI receptionist never do unsupervised?
It should never quote a price, promise a deadline, or confirm availability it has not verified against your real calendar or price list, and it should never guess at an answer it does not know. Review transcripts and bookings daily for the first month, and route payments, complaints, and urgent calls to a person.
- BT launch announcement via WebWire (October 2, 2026) and Mobile News report on BT’s AI Receptionist launch (October 2026).
- BT/Censuswide survey (late September 2026, 1,500 UK SME decision makers + 2,000 UK adults).
- Upfirst, The State of Voice AI for Small Businesses 2026 (October 14, 2026).
- ToolChase, Best AI Receptionists and Phone Agents in 2026 (September 2026 pricing).
- BBC News, Business Daily: Can businesses trust AI? (November 2, 2026).
- BBC News, GP surgery AI receptionist report (October 2026).
Was this article helpful?