An AI chatbot is worth it when the enquiries it captures while you are busy, closed or on the tools are worth more than it costs to run. For most NZ service businesses that is a question you can answer with four numbers you already have: how many enquiries you miss, how many of those would have become jobs, what a job is worth, and what the chatbot costs. We sell chatbots, so read this as a method for checking our claim rather than a reason to believe it. It also covers the situations where the honest answer is no.
What does an AI chatbot on a website actually do?
A modern website chatbot is trained on your own content: your services, prices, service areas, hours and the questions you get asked every week. When a visitor asks something, it answers from that material, in plain language, at any hour. When the visitor is ready, it asks for a name, a phone number and enough detail about the job to be useful, and sends that to you. On some setups it can book a slot in your calendar. What it cannot do is quote a complex job accurately, promise things you have not told it, or replace the conversation that wins the work; its job is to make sure that conversation happens instead of going to the next business in the list.
You will find no statistics in this article about how much chatbots increase conversions, because the honest figures depend entirely on your business, and the ones quoted in sales pitches rarely come with a source. The method below is worth more than any of them.
Where do the missed enquiries come from?
Before you can value a chatbot you need to know what it is catching. For most small businesses the leaks are predictable:
- After hours. People research tradies, clinics and services in the evening and at weekends. A form they fill in at 9pm often never gets sent; a question they cannot get answered sends them elsewhere.
- While you are working. A plumber under a sink and a hairdresser mid-colour do not answer the phone. Some callers leave a message; many call the next name.
- The question before the enquiry. "Do you do Waiheke?" "Is it a callout fee plus hourly?" "Do you take Southern Cross?" If the website does not answer, the visitor either asks or leaves, and most leave.
- The channels that have closed. Google removed the chat feature from Business Profiles: as of 31 July 2024 chat and call history are no longer available, and Google points customers to your website links instead. If you want a chat channel, it now lives on your site.
You probably already have a feel for which of these applies to you. The next step is turning that feel into a number.
How do you work out whether a chatbot pays for itself?
Four numbers. Write them down honestly; rough is fine.
- Missed enquiries per month. Count for a fortnight: after-hours calls that went to voicemail, missed calls you never returned, form submissions that arrived at odd hours and went cold, and website visits at times you were closed if your analytics show them. Double it for the month.
- Your conversion rate. Of the enquiries you do answer, how many become paid work? If you do not know, look at last month's enquiries and last month's jobs.
- Average job value. What a typical first job is worth to you, not your biggest.
- Monthly cost. The plan fee plus AI usage. Our numbers are in the next section.
Then: missed enquiries, multiplied by your conversion rate, multiplied by average job value, is the revenue at stake each month. A chatbot will not recover all of it, so take a fraction you believe. If that fraction is comfortably above the monthly cost, it pays; if it is below, it does not, and no amount of enthusiasm about AI changes that.
A worked example, with made-up numbers purely to show the arithmetic. Suppose a sole-operator electrician counts eight missed enquiries a month, converts one in four answered enquiries into a job, and a typical first job is worth $350. The revenue at stake is 8 × 0.25 × $350, which is $700 a month. If a chatbot recovers even a quarter of that, $175, it covers a $99 plan with AI usage on top and leaves a margin. If the same electrician missed two enquiries a month, the sums say wait. Put your own numbers in; the example proves nothing about your business.
What does a chatbot cost with Go Digital?
We keep it simple and we publish it. The chatbot is included on our Growth plan at $99 a month and, in an advanced form, on Pro at $179, both with $0 upfront and no lock-in; those prices cover the whole website, not the chatbot alone, so a business already on Growth pays nothing extra for the plan side. AI usage, meaning the cost the AI provider charges per conversation, is billed separately: you bring your own API key, or we pass the usage through at cost. We will show you the provider's current pricing before you commit, rather than quoting a figure here that would be out of date within months. Our chatbots and voice agents run on KiwiAgent, which is our own sister product rather than a third-party tool we resell; we say so on our straight answers page because you should know who you are buying from.
When is a chatbot not worth it?
We would rather tell you now than after you have paid for one:
- You already answer everything. If someone picks up every call and replies to every form within minutes, there is little to catch.
- Enquiry volume is tiny. A business with a handful of enquiries a month has a marketing problem, not a response problem. Fix visibility first; our guide to websites missing from Google is the place to start.
- Every job needs a human to sell it. Complex, high-value work where the first conversation is a consultation rather than a booking gets less from a bot, though it can still capture the contact details.
- The website is the problem. A slow, confusing site with no clear services loses visitors before they would ever open a chat. A chatbot on a bad website is a receptionist in an empty shop.
- You have nothing to train it on. If your prices, areas and policies are not written down anywhere, the bot cannot answer accurately. Write them first, and you will find the website improves too.
What makes a chatbot actually capture leads?
- It is trained on your real content, not a generic script. If your callout fee is $120, it should say so, and it should never invent a price you have not set.
- It asks for a phone number early. The purpose is a lead, not a long conversation. Two or three questions, then contact details.
- It says it is a bot. People do not mind talking to a bot; they mind being tricked. Honesty here also keeps you on the right side of fair trading expectations.
- It hands over cleanly. When it cannot help, it says so and gets the enquiry to you, rather than looping.
- It knows it is in New Zealand. Time zone, spelling, suburbs, GST, the fact that "sparky" is a job title.
- You read the transcripts. Weekly, for the first couple of months. Every question it could not answer is a gap on your website.
How do you measure it after launch?
Decide the measure before you switch it on, or you will argue about it later. The useful ones are conversations started, leads captured with a phone number, and jobs won from those leads, compared with the month before. If your analytics are set up for it, conversations and lead submissions should be tracked as events alongside form submissions and calls, so you can see the chatbot's share rather than guessing. The monthly performance report on our Growth and Pro plans includes this; our guide to telling whether your website is working explains how to read the rest.
Do this today
- Start a fortnight's tally of missed calls and after-hours enquiries, on paper if necessary
- Work out last month's enquiries-to-jobs ratio from your invoices
- Write down your prices, service areas, hours and the ten questions you are asked most; you need this list whether or not you buy a chatbot
- Do the four-number sum above with your own figures
- If it pays, decide the success measure before switching anything on
Where we come in
If the sum says yes, the chatbot is already part of our Growth and Pro plans, trained on your content during the build and connected to your analytics from day one. If the leak is phone calls rather than web chat, the same platform offers an AI voice agent that answers when you cannot, takes the details and sends you a summary. And if the sum says no, we will tell you that too; the free mock-up and the free website audit both cost nothing, and a business that does not need a chatbot usually needs one of those first.
Frequently Asked Questions
How much does an AI chatbot cost for a small business in NZ?
With Go Digital the chatbot is included in the Growth plan at $99 a month and in advanced form on Pro at $179, with $0 upfront; those prices cover the whole website. AI usage, the per-conversation cost charged by the AI provider, is billed separately: bring your own API key or pay usage at cost. Other providers price chatbots as standalone subscriptions, so compare the total including usage.
Will an AI chatbot annoy my customers?
A badly built one will: pop-ups that interrupt, scripts that loop, and bots pretending to be people. A good one is unobtrusive, says it is a bot, answers from your real content, asks for contact details quickly and hands over when it cannot help. Customers who arrive after hours are generally glad to get an answer at all.
Can a chatbot book appointments?
Yes, when it is connected to your calendar. It can offer available times and confirm a booking, or simply collect the preferred time with the enquiry for you to confirm. Which approach is better depends on how much your jobs vary; a salon can book directly, a builder usually wants to call first.
Can AI answer my phone as well as my website chat?
Yes. An AI voice agent answers calls when you cannot, asks what the caller needs, takes their details and sends you a summary, so a missed call becomes a lead instead of a lost one. Go Digital's voice agents run on KiwiAgent, its own sister product, and are included on Growth and Pro with AI usage billed separately.
Do I need a chatbot if I already use Facebook Messenger?
Messenger only helps people who are already on Facebook and already talking to you. Most people looking for a service search Google and land on a website, and a chatbot there catches the enquiry at that moment. Keep Messenger for existing customers; use the website for the ones who do not know you yet.
The bottom line
A chatbot is not an AI decision; it is an arithmetic decision. Count what you miss, value it honestly, compare it with the cost, and act on the answer either way. If you want us to run the numbers with you, get in touch; the conversation is free and it ends with a yes or a no.