Straight answers, before you call.

Most small businesses do not need AI. They need two or three repeated jobs taken off the desk, and AI happens to be a cheap way to do that now.

The word matters less than the job. If nothing in your week is repetitive, predictable and costing you money when it slips, you do not need any of this and nobody here will pretend otherwise.

No. Small is where it helps most, because there is nobody to hand the work to.

A large company answers its phone by hiring a receptionist. A three-person shop answers it by stopping work, or by not answering it at all. That gap is the whole reason this exists.

Businesses where the owner is busy with their hands while the phone rings. Trades, salons, clinics, restaurants, shops and anyone taking bookings.

The common pattern is simple: every enquiry is worth real money, and every enquiry arrives while you are in the middle of something else.

Do not change what works. Paper is not the problem.

The thing worth looking at is the call that came in while you were up a ladder, and where that customer went instead. Nothing on this page asks you to change how you do the work.

Usually the missed call. It is the quickest to build, and it is the one where you notice the difference within a week.

Starting small is also the point. You find out whether this suits your business before spending on anything bigger.

Three questions. Does it happen every week, does it follow the same steps every time, and does it cost you a customer or money when it is late?

Three yes answers means it is probably worth it. One yes means leave it alone. That test is applied before Flint quotes anything, and it rules jobs out often.

Anything that needs judgement. Pricing an unusual job, an unhappy customer, a complaint, anything to do with safety or money moving.

A system should hand those to you faster, not answer them. The right use is clearing the routine off your desk so you have time for the calls that need you.

No. That is a sales line.

Businesses in Halifax have been running well for decades without any of this and will keep doing it. The realistic risk is smaller and duller: the competitor who answers a text at seven in the evening gets the job you never heard about.

Ordinary automation follows rules somebody wrote in advance. If this, then that. AI handles the parts where the input is messy.

A booking rule cannot read "any chance you could squeeze me in Thursday after 4". AI can read it, turn it into a time, and hand the rest to ordinary automation. Most good systems are mostly ordinary automation with AI in the one spot that needs it.

Because ChatGPT is a tool you have to remember to open. A system runs whether or not you think about it.

Nobody gets value from a chat window at the end of a long day. The value comes from something that answers a text at 9pm, puts a booking in your calendar and tells you in the morning.

When a call to your number goes unanswered, the caller gets a text within seconds saying you are on a job and asking what they need.

They reply by text, which most people find easier than leaving a voicemail. The system can take the details, answer the ordinary questions and book a time, or simply collect the information and pass it to you. You get the whole conversation on your phone. Nothing about your calls changes when you do answer.

Yes. The systems Flint builds say so, in the first message, in plain words.

Pretending otherwise is how a business loses trust in one screenshot. Said openly it reads as organised, not cold, because the customer got an answer instead of silence.

It stops trying and gets you, immediately. A system that keeps chatting to somebody with a burst pipe is worse than no system at all.

How that works is decided with you before it goes live, in plain language. You name the words and situations that count as urgent for your trade, a flood, no heat, a lock-out, a child, and any message matching them skips the whole conversation. The caller is told plainly to phone you now, your real number is put in front of them, and your phone rings or buzzes straight away. When the system is unsure, it is built to treat it as urgent and hand over. The failure everyone hears about is a customer trapped in a loop with a machine, and the fix is deciding the handover rules before anything is switched on, not afterwards.

A text system hands over by telling the customer to call you and alerting you at the same time. A system that answers the phone in a voice can transfer the call to a real number.

Voice answering is the hardest of these to do well and Flint will not promise you a date for one on a first call. Text is where most small businesses get the result, and it is far more forgiving when something goes wrong. Whichever is built, there is always a route to a human, and you decide what it is.

Nobody can tell you that number from the outside, and anyone who quotes you one is guessing.

You can find it out in about a week. Count the calls you did not answer, then count how many of those people called back. The ones who never called back went to whoever picked up next. That number is yours and it is the only one worth basing a decision on.

Usually yes. Google Calendar, Outlook and most booking tools allow it.

It reads your real availability, so it cannot offer a time you are already busy. Before anything is built, your calendar is checked to confirm it will connect, and you are told straight away if it will not.

In most cases your number stays exactly as it is. Nothing is ported and nothing about answering calls yourself changes.

Some setups add a second number that forwards to yours. If that is the case for your setup you are told before anything is built, because a business number on a van and a sign is not a thing to change casually.

It books into your real calendar, so a time you are already busy cannot be offered.

For the first stretch after anything goes live you can have every booking come to you for a quick yes before it is confirmed. Plenty of owners keep it that way permanently. Every booking is also visible to you the moment it happens, so a mistake is caught in minutes rather than on the day.

Yes, and for most businesses that is now the bigger pile.

Texts, messages on your Facebook page and Instagram messages can all run through the same system, so an enquiry gets the same answer no matter where it arrives.

Often yes. Most common trade and booking software has a way to connect, and where it does not there is usually a workable route around it.

The honest part is this: your software is checked first, and if it will not connect properly you are told before you pay for anything, not after. No system is worth buying twice.

They come from public listings: map listings, directories and public business records, filtered down to your area and your trade.

They are not people who asked to hear from you. They are a list of businesses or households worth approaching, which is useful if your trade sells to other businesses and much less useful if you sell to walk-in customers. If your business is the second kind you will be told to spend the money elsewhere.

It is spam when it is sent to everybody. It is not when the list is small, the message is about that specific business, and a reply saying stop is honoured permanently.

Canadian law sets real rules for commercial email, including who you can contact and giving a working way to unsubscribe. Anything Flint builds follows them, and anything that would break them does not get built.

A Facebook page is enough to be found. It is usually not enough to be believed.

People check a business before spending real money, and a page with three posts from last year answers a different question than a site with your services, your area and your phone number. It also matters more now that people ask an assistant for a recommendation, because those tools read websites far more readily than social pages.

The ones that jump out after two seconds and ask how your day is going, yes.

A quiet one that sits in the corner and answers what you charge for a call-out, whether you cover Dartmouth and how soon you can come, does not. If a chat box on your site would have nothing useful to say, it should not be there.

Only if it is writing from your own material. Your photos, your jobs, your prices, your words.

Given nothing to work from it produces the bland posts everyone recognises instantly. The version that works is you sending three photos from a job and getting back a post you can approve or change in ten seconds.

Those tools read the ordinary web. What helps is having a real website that plainly says what you do and where, a Google Business Profile that is filled in and matches it, and reviews.

Nobody can buy a place in those answers and anybody selling you one is selling nothing. The businesses that get named are the ones whose details are consistent everywhere and easy to read.

Almost every local business needs a website and does not need an app.

People will not install an app to book a haircut. An app earns its cost when your own staff need a tool on the job that does not exist yet, which is a different conversation from getting customers.

Most of it is stored outside Canada, usually in the United States, because the services these systems are built on are American. You are told which services hold what before anything is set up.

That is the same position as most software a Canadian small business already uses, including common email, booking and card-payment tools. If storing customer information in Canada matters for your work, say so at the start, because it narrows what can be built and is far easier to plan for than to fix later.

Handling customer information with software is normal and legal. The rules are about how you do it, not whether you use AI.

In practice that means collecting only what you need, telling people what you collect, keeping it secure and not using it for something they would not expect. Flint builds to that standard and will tell you where a request crosses a line. That is a description of how the work is done, not legal advice, and for anything sensitive such as health records you should talk to a lawyer.

Every system Flint builds says so anyway, so the question rarely arises.

Canadian rules in this area are still being written and vary by what you do. Saying it plainly costs you nothing, protects you if the rules tighten, and customers respond better to it than to being fooled.

It depends on the service and the plan, which is exactly why it is checked before anything is built.

Business plans on the major AI services do not train on what you send them by default. Free and personal plans sometimes do. You are told which service a system uses and what its policy says, so the decision is yours rather than a surprise.

You do. Wherever a service allows it, the accounts are opened in your business name, so your data and your customer list stay in accounts you control.

If you walk away you keep the accounts, the data and the working system. Nothing is held hostage, and there is no version of this where leaving means losing your customer list.

Probably not for a job this size, and it is better to hear that now than after an application.

The federal digital adoption program that used to cover this closed to new applications in 2024. The provincial innovation rebate is built for capital projects far larger than a small business automation. The Atlantic Canada Opportunities Agency does fund AI work in the region, but it is aimed at substantial projects and runs through their offices, so a phone call to them is the honest first step if you want to try. Plan on paying for a small system yourself.

Yes. Kade Lavoie builds these himself and lives in Halifax, so sitting at your counter on a weekday is a real option rather than a promise in an email.

It helps more than people expect. Watching how the work actually gets done in your business usually changes what gets built.

Yes. The South Shore, the Valley, Cape Breton, New Brunswick and Prince Edward Island are all served, and most of the work is done remotely anyway.

Distance changes the in-person part, not the system. Further out it is usually one visit rather than several, with the rest by phone and screen share.

Yes, and it is worth saying at the start. Seasonal businesses should not pay monthly for a system that is idle half the year.

Most of these can be paused in your off season and switched back on before you open, or set to simply take a message and tell people when you reopen. That is a decision made when it is built, not something you have to fight for later.

You call or text Kade at the number on this site. There is no ticket queue and no support address that nobody reads.

Anything that stops customers reaching you is treated as urgent and dealt with the same day. Smaller things get picked up in the ordinary course of the week.

You own it and there is no contract keeping you. The accounts are in your name wherever a service allows it.

The realistic tie is that somebody has to look after it, and most owners would rather that be the person who built it than themselves. That is a choice you keep making month to month, not one you sign away at the start.

A small system is usually days rather than weeks. Something wired into several tools you already use takes longer.

You get a date with the quote, before any work starts. If something is going to be late you hear it from Flint rather than finding out by waiting.

Less than most owners fear. One proper conversation, then short answers to specific questions as they come up.

The first conversation is the one that matters, and it is about how your business actually runs rather than anything technical. After that you are asked for the details only you have, such as your prices, your hours and what you want said to a customer at nine in the evening.