AI Solutions for Small Business: What Actually Works in 2026
Most small business owners I talk to have already tried AI. They opened a chat window, asked it to write a quote email, thought it was clever, and then went back to work. Nothing changed in the business.
The gap is not the technology. The gap is that a chat window is a tool you have to remember to use, and you are busy. The AI solutions for small business that actually shift the numbers are the ones that run without you: they sit on your phone line, your website and your inbox, and they do the same job every time whether or not you thought about them that morning.
This is the overview I give to owners before we talk about anything specific. What the technology genuinely does today, what it costs in real money, what it still does badly, and how to choose the one project worth starting with.
The AI solutions for small business that reliably work today#
Ignore the demos. In a two-to-fifty person service business, there are three places where AI produces money rather than novelty, and they map onto the three moments where small businesses leak the most.
The first is answering enquiries. Calls, web chats, texts and social DMs, at nine at night and on Sunday afternoon, in your business's voice, with your prices and your service area. An AI agent answers the question, books the job into your calendar, and passes anything unusual to you with the full conversation attached. This is the highest-return use for most service businesses because the loss it fixes is invisible: a missed call does not appear anywhere in your accounts, it just quietly becomes someone else's job. Start with the complete guide to an AI receptionist if the phone is your bottleneck.
The second is converting the traffic your website already gets. Most small business sites are brochures. Someone lands, reads two paragraphs, cannot find a price, and leaves. A site with an AI layer answers their actual question in the moment, works out whether they are a real buyer, and offers a booking slot instead of a contact form. Same traffic, different outcome. That is the subject of what makes a website smart.
The third is keeping content moving. Not ghostwriting your brand, but removing the friction that makes you stop posting by week three: ideas from your own jobs, drafts in your voice, scheduling, and fast replies to comments. AI is genuinely good at the first and last of those and mediocre in the middle, which I go into in AI social media management.
Everything else being sold to small businesses right now is either a thinner version of one of these or a solution looking for a problem.
What each one costs, honestly#
Prices vary by who builds it and how much of your business it needs to know. These are the bands I see in New Zealand, in NZD per month, excluding GST.
| Job | Self-serve tool | Configured system | What drives the cost |
|---|---|---|---|
| Answering calls and chats | $50 to $200 | $400 to $1,200 | Voice minutes, calendar integration, number of escalation rules |
| Website conversion layer | $50 to $150 | $300 to $800 | Depth of qualification, CRM connection |
| Content and social | $30 to $120 | $400 to $1,000 | How much human editing stays in the loop |
Setup fees for a configured system usually land between $1,000 and $5,000 depending on how much of your process has to be written down before anything can be built. That number surprises people, and it should not. The AI is cheap. Working out what your business actually promises customers is the expensive part, and it is expensive whether or not you automate afterwards.
The self-serve column is not a trap. For a lot of businesses it is the correct answer, and I say so often enough that it costs me work. A single-operator business doing twenty enquiries a week does not need a custom build. The moment to move up a band is when the cheap tool starts giving answers you have to apologise for.
Where AI still gets it wrong#
I would rather you go in knowing the failure modes than discover them in front of a customer.
It does not know what it does not know. An AI agent given a vague brief will answer confidently and wrongly about a price you never gave it. The fix is a tight knowledge base and an explicit instruction to escalate rather than guess, which is why what you feed the agent matters more than which model is under the hood.
It handles emotion badly. A customer who is angry, grieving or spending a lot of money wants a person. Good systems detect that quickly and hand over cleanly, and the handover is where most implementations fail. A customer who has to repeat their story to the human is worse off than if the AI had never answered.
It cannot fix a broken process. If your quoting takes nine days, an AI that responds in nine seconds just gets the customer to the queue faster. Automating a bad process makes the bad process more efficient, nothing more.
It will not replace your judgement on price or on who you want to work for. Everything upstream of those decisions is fair game.
Picking the first project#
The businesses that get value from AI pick one problem and finish it. The ones that do not pick four, half-configure all of them, and conclude the technology does not work.
Score each candidate problem against four questions, and be honest about the answers.
- How many times a week does this happen? Under ten and the savings will not cover the effort of setting it up, no matter how annoying the task is.
- Does it have a right answer you could write down? If two of your staff would answer the same customer question differently, the AI has nothing to learn from.
- What happens if it gets one wrong? A misfiled email is recoverable. A wrong medication instruction or a wrong quote on a $30,000 job is not, so those stay human or stay supervised.
- Who checks the output in week one? If the answer is nobody, the project is already finished.
For most service businesses the answer that survives this test is enquiry response, because the volume is high, the answers are consistent, and a mistake usually costs you a follow-up call rather than a customer. I have written the fuller version of this scoring exercise in how to choose your first AI project.
Work out whether the numbers justify it before you spend#
Do this on paper, with your own figures, before anyone quotes you.
Take your inbound enquiries per week. Estimate what share currently go unanswered or get a reply the next day. Take the proportion of those that would have turned into work, and multiply by your average job value. That is your monthly leak. Compare it against the monthly cost from the table above, and you have a payback period.
Take a plumbing business fielding 40 calls a week and missing roughly a third of them. If a fifth of those missed calls would have become jobs at an average of $480, the arithmetic gets uncomfortable fast, and the monthly cost of the system stops being the main number in the conversation. Run your own version of that calculation in working out the ROI of AI automation, including the case where it does not pay off, because that case is real and worth finding early.
The other half of the sum is the one nobody does: what does it cost you to keep the status quo, in your own evenings? A lot of owners are the after-hours answering service and have stopped counting the hours.
Baseline your numbers before you launch anything#
The single most common regret I hear is that a business installed something, felt busier, and could not prove it worked. Before you switch anything on, record how many enquiries you get in a week, how many you answer within an hour, and how many become quoted work. Two weeks of that is enough. Without it, you will be arguing about a subscription in month four with no evidence either way. The five metrics worth tracking covers what to capture and how.
What I would do this week#
If you are starting from nothing, spend an hour on this and no money at all.
- Count your missed calls for the last fourteen days from your mobile log or phone system.
- Write down the twenty questions customers ask you most, with your real answers and prices.
- Pick the single moment where you lose people most reliably, and decide whether it is the phone or the website.
That third document is the asset. Whether you build something yourself, buy a tool for $80 a month, or work with someone like me, the businesses that write down their own answers first get a system that sounds like them. The ones that skip it get a generic bot and blame the technology.
If you are in Auckland and want the local picture on costs and labour comparisons, the Auckland guide goes into what the numbers look like here specifically.
Common questions
What can AI actually do for a small business right now?
It answers phone calls, chats and DMs in your business's voice, books jobs into your calendar, qualifies website visitors before they reach you, and drafts and schedules content. These are narrow, repeatable jobs with a clear right answer, which is where the technology is dependable.
How much does AI cost for a small business?
Self-serve tools start around $50 to $150 a month per job. A configured system built on your prices, hours and escalation rules usually runs $300 to $1,200 a month with a setup fee. Custom builds start in the low thousands and only make sense for unusual workflows.
How long does it take to get an AI system working?
A self-serve tool is running the same day but needs a few weeks of correction before it is trustworthy. A configured system typically takes two to four weeks from first conversation to handling live customers, with the bulk of that time spent gathering your prices, policies and escalation rules.