Most owners asking whether an AI employee is worth it are really asking something narrower: will this thing answer my customers properly, or will it embarrass me and add another subscription I forget to cancel? That is a fair worry. Plenty of businesses have bought software that sounded clever in a demo and did nothing useful on a Tuesday afternoon.
The honest answer depends less on the technology and more on the shape of your enquiries. A website assistant is very good at some jobs and pointless at others. The ten questions below are the ones we would ask if we sat down with you and your inbox for half an hour.
Score each one as you go. At the end there is a simple way to add it up, and a clear description of what a "no" looks like.
Questions about your enquiries
1. Do the same questions arrive again and again?
Open your inbox or contact form history and read the last 30 enquiries. If you can sort most of them into a handful of piles (prices, availability, service area, how long something takes), that repetition is exactly what an assistant absorbs. If every enquiry is a bespoke conversation, it will mostly be collecting names for you.
2. How many enquiries do you get, and how many go unanswered for hours?
Volume matters less than delay. Ten enquiries a week that each sit for a day can cost more than fifty that are answered within minutes. If you have not counted this, what missed enquiries actually cost walks through a simple tally.
3. When do enquiries arrive?
Look at timestamps. Evenings, weekends and lunchtimes are when a small team is least able to respond and when an always-on assistant does the most obvious good.
Questions about your content
4. Could a new starter answer common questions from your website alone?
This is the best single test. If a sensible temp could read your site and handle the easy enquiries, an assistant trained on the same pages can too. If your site says "contact us for details" on every page, it cannot.
5. Do your answers exist somewhere, even if not online?
Price sheets, onboarding PDFs, old email replies and staff notes all count. An assistant can learn from uploaded documents and question and answer pairs, not just the website. What it cannot do is know things nobody has written down.
6. Are your prices and policies stable enough to publish?
If you quote every job from scratch and would rather not give ranges, the assistant will spend its time saying "someone will get back to you". That still captures the lead, but it is a smaller win.
Questions about your team
7. Who will pick up the leads and tickets it creates?
An assistant that captures a booking request at 10pm is only useful if someone reads it at 8am. Name the person now. If nobody owns it, the answer is no until somebody does.
8. Is there a person who can take over live when it matters?
Not essential, but valuable. Some enquiries need a human within minutes, and a clear handover route makes the assistant safer to deploy.
If you run a clinic, a legal practice or anything touching money or safety, treat this question as close to essential. The assistant should give general information only, and anything clinical, legal, financial or urgent needs a qualified person or the right emergency number. Knowing who that person is, and how the visitor reaches them, is part of being ready.
9. Will someone spend 20 minutes a week reviewing what it could not answer?
Unanswered questions get recorded as knowledge gaps. Filling them is how the assistant improves. Without that habit it stays exactly as good as day one.
The question most people skip
10. What happens to an enquiry today when nobody answers?
Be precise. Does the visitor fill in a form, ring back tomorrow, or quietly try the next business in the search results? If the honest answer is "they wait patiently and it is fine", you have less to gain than you think.
Adding up your score
Give each question a 0, 1 or 2 based on how strongly it applies.
| Score range | What it usually means |
|---|---|
| 0 to 7 | Probably not worth it yet. Fix content or ownership first. |
| 8 to 13 | Worth a small test on one or two busy pages. |
| 14 to 20 | A strong fit. Plan a proper setup and a named owner. |
Here is a worked example you can swap your own numbers into. Say a small cleaning company gets 40 enquiries a week, and roughly 15 arrive after 6pm or at weekends. About 25 of the 40 are variations on price, area covered and availability. The website lists prices and postcodes, the office manager already checks the inbox each morning, but nobody reviews anything weekly. That business might score 2, 1, 2, 2, 2, 1, 2, 2, 1, 1 for a total of 16. The one weak point, the weekly review, is cheap to fix.
If you prefer money to scores, the chatbot ROI calculator lets you plug in enquiry volume and value.
When the answer is no
It is worth being blunt about the businesses that should not bother, at least for now:
- One-off, high-touch sales: if every client needs a long discovery call before anything useful can be said, the assistant is little more than a contact form.
- No written knowledge at all: it cannot invent your policies, and it should not try.
- No owner for follow-up: captured leads that sit unread are worse than a form, because the visitor thinks they have been heard.
- Tiny, patient demand: a handful of enquiries a month from customers who are happy to wait does not need 24/7 cover.
There is a longer version of this argument in when not to buy an AI receptionist, and it is worth reading before you sign up for anything.
What to do this week
Pull your last 30 enquiries and mark each with the time it arrived and whether it was a repeat question. That hour of sorting tells you more than any sales call.
If you scored in the middle band, do not plan a big rollout. Put an assistant on your single busiest page for two weeks using a free plan, try it on your own site first, and read every conversation. If it answers well and captures enquiries you would otherwise have missed, widen it. If it mostly says "I don't know", your content needs work before any tool will help.