The numbers, and whether they survive.
Before we built any of this we went through 41 companies selling something like it, and graded the twenty statistics this industry quotes most. Six hold up. Ten are usable if you say where they came from. Four should never be repeated by anyone. We publish the lot, including the ones that make our own pitch harder.
official statistics or independent academic research with a stated method
real research, but vendor-published, dated or paywalled. Always attributed
no traceable primary source, or so old that repeating it misleads
pricing pages read directly, because third-party pricing articles are often stale
Every number we could have used, graded
Filter it. If you are buying automation from anyone, including us, ask where their numbers came from and see whether the answer is on this list.
Safe to publish 6 of 20
Official statistics or independent academic research with a stated method. Quote these freely.
Adoption rose eight percentage points in a year, and reaches 49% among firms with 250+ staff. Yet the number of AI technologies per adopting firm moved only from 1.4 to 1.6. Nearly everyone has tried a chatbot; almost nobody has a workflow that runs without them.
Office for National Statistics, Artificial intelligence in UK businesses: 2023 to 2026 · ons.gov.uk · 2026
Rising to 24% above a 50% response rate, with average ratings climbing from 3.81 to 4.04. Returns diminish past about 40%, so there is no need to answer everything.
Cornell Hospitality Report, Anderson & Han, When Online Reviews Meet Hotel Response · sha.cornell.edu · 2016
At the same occupancy. The commercial argument for a review workflow in a single sentence.
Cornell Hospitality Report, Anderson · travelweekly.com · 2012–16
Up from 12% the year before. Among 18 to 34 year olds it is 28%, against 1% of over-55s. From a nationally representative survey of 5,000 consumers.
CGA and Zonal, GO Technology · cgastrategy.com · 2024
Anxiety rose from 73% the previous year, and 46% say they simply lack the knowledge to use it well. This is the objection our two-week human review period exists to answer.
Federation of Small Businesses · smeweb.com · 2026
Useful for showing a trades or property audience how far ahead other sectors already are.
Office for National Statistics · ons.gov.uk · 2026
Use with care 10 of 20
Real research, but vendor-published, dated or paywalled. Fine with attribution, never as a headline.
The slowest agent in the study took the equivalent of five working days. Only around 40% of sales enquiries were answered within the hour, against roughly 70% of rental ones. Research by a company that sells to agents, but a real mystery shop with a stated method, and the most on-target figure available for this market.
Homeflow, reported by Estate Agent Today · homeflow.co.uk · 2023
From the same Homeflow mystery shop. Strong, and vendor research. We say so in the same breath.
Homeflow · homeflow.co.uk · 2023
Widely quoted across the property sector. Attribute it to industry reporting rather than presenting it as independent research.
Property sector reporting · agentzap.ai · 2026
A real study of 15,000 leads and over 100,000 call attempts. But it is from 2007, it was published by a company selling sales software, and it is very commonly misattributed to Harvard Business Review. Attribute it correctly or leave it out.
MIT / InsideSales, Dr James Oldroyd · insidesales.com · 2007
A credible source, but a US sample and the full text sits behind a paywall.
Harvard Business Review, The Short Life of Online Sales Leads · hbr.org · 2011
Widely reported and directionally believable, but it originates with a company selling hotel voice AI. Fine as what the industry reckons; not as a fact.
Canary Technologies, via Hotel Dive · hoteldive.com · 2026
A vendor claim, originally in dollars. A useful shape for a calculator, not a citation.
Loman.ai · loman.ai · 2026
Modelled by extrapolation from survey data rather than measured. Always say estimated.
Zonal and CGA, via The Caterer · thecaterer.com · 2021–24
A vendor survey across three countries, so directional rather than authoritative. It does support the case for publishing a security page.
Xero · xero.com · 2026
A competitor's case study, not ours. Included because it shows what a credible result in this market looks like, and sets the bar our own first case study has to clear.
HiJiffy customer story · hijiffy.com · 2025
Never use 4 of 20
No traceable primary source, or so old that repeating it misleads. We list them so you recognise them.
Traces to a single press release from 2014 by a company that sells phone systems. Twelve years old. Almost every article you will find about the cost of missed calls leads back to this one document. We will not build a calculator on it and neither should anyone else.
BT / Populus press release · mynewsdesk.com · 2014
Circulated endlessly across answering-service marketing. No primary study exists that anyone can locate.
No traceable source · — · —
Vendor marketing. The direction is widely accepted; the number is not verifiable anywhere. If the point is worth making, make it qualitatively.
Booking platform marketing · unverified · 2024–25
A press release wire item based on 142 respondents with no published method.
openPR wire release · openpr.com · 2025
Nine pages blocked automated reading during this research, among them the Information Commissioner's Office, Estate Agent Today, Moneypenny and Popmenu. Anything sourced from those was taken from search summaries and is flagged accordingly. Prices and claims about competitors were correct in September 2026 and change often.
Because the alternative is asking you to take our word for it
Every company in this market quotes the same handful of alarming statistics at small business owners. We went looking for where they came from. Several came from nowhere at all.
We would rather hand you a shorter list of numbers that hold up, and be the supplier who tells you which ones don't. It is the same instinct behind the rule that the AI never invents a price: if we won't let a workflow guess at your customers, we shouldn't guess at you either.
Next stepTwenty minutes on the phone. Then a straight answer.
Tell us what you keep doing by hand. We'll tell you which workflow pays back first, what it costs, and whether it's worth doing at all. If it isn't, we say so.