Ask five shops what a custom tool costs and you'll get five numbers a decimal place apart. That's because most of them are quoting a different thing than the one you're picturing. A weekend script that texts back missed calls and a full platform your whole crew runs on both get called "custom software," and the gap between them is roughly the gap between a truck and a fleet. Here is how the money actually breaks down, in Canadian dollars, from someone who builds and runs these systems in Edmonton.
Every Honest Quote Has Two Numbers
There's a one-time build fee and an ongoing monthly. If a quote only shows you one of them, the other is hiding somewhere. Software isn't a fence you install and walk away from. It runs every day, talks to other services that change their rules without asking, and needs someone on the hook when it breaks at 7am on a Monday.
For a single focused automation, the build fee usually lands somewhere in the $3,000 to $15,000 CAD range, and the monthly to host and operate it runs from under $200 up to a few hundred (illustrative, and genuinely dependent on scope). A real multi-part platform your team logs into daily is a different animal: North American benchmarks for that kind of build sit in the tens of thousands and climb past $100k for anything ambitious. Small businesses rarely need that on day one, and shops that push you toward it on day one are the ones to watch.
The monthly is not a subscription tax. It covers hosting, the AI usage itself when there's a model involved, security patches, and the human who fixes things. Skip it and you've bought a car with no oil changes. It runs great right up until it doesn't.
What Actually Moves the Price
The feature list is almost never what drives cost. Integrations and edge cases are. Connecting one clean system like Stripe or Google Calendar is cheap and well-documented. Connecting to an old QuickBooks setup, a booking tool with no real API, and a spreadsheet three people edit by hand is where the hours pile up, because now someone has to handle every way those systems disagree with each other.
The second driver is how wrong the software is allowed to be. A tool that drafts a quote for a human to approve is forgiving. A tool that sends the quote to the customer and books the job with zero human in the loop has to be right nearly every time, and that reliability is most of the engineering. Decide up front where a person stays in the loop; it's the single biggest lever on your bill.
- Clean modern integration (Stripe, Google Calendar, Jobber): low cost, mostly predictable
- Legacy or API-less system, or hand-maintained spreadsheets: the real budget sink
- Draft-for-human-review: cheaper, safer, ships faster
- Fully autonomous send/book/charge: far more testing, far higher price
- Data you own and can export vs. locked inside a vendor: ask before you sign anything
A Worked Example: The Missed-Call Problem
Say you run an Edmonton trades outfit and you miss 12 calls a week between January's cold snaps and being up a ladder. Call-tracking data consistently shows most callers who hit voicemail just dial the next business on the list, so treat a missed call as a lost job, not a callback later. If your average job is $650 and you close even one in four of those callers, the missed calls quietly cost you roughly $2,000 a week.
A tool that texts every missed caller within seconds, asks two qualifying questions, and drops the answer into your calendar is a modest build. Illustratively, call it $4,000 to $8,000 CAD to build and around $150 to $300 a month to run, including the SMS and AI costs. Against $2,000 a week of leaking work, it pays for itself inside the first month and keeps paying. That's the math that justifies a build. If your version of this loses you $80 a month, don't build anything.
The Costs That Never Make The Quote
Hosting is small but real, often $20 to a couple hundred a month depending on traffic. When AI is involved, the model charges per use, and this one surprises people: a chatbot answering a few hundred questions a month costs pocket change, but one summarizing long documents all day can run into real monthly money. Ask any vendor to estimate your token usage out loud, and be skeptical of a flat 'AI included' line with no numbers behind it.
Then there's maintenance, which the industry roughly pegs at 15 to 20 percent of the build cost per year, and the quieter cost of internal ownership. Someone on your side has to answer the tool's questions, handle the exceptions it kicks out, and tell you when a process changed. Budget for that person's time. A system nobody owns internally slowly rots no matter how well it was built.
When Not To Build It
Plenty of the time the right answer is don't. If a $40-a-month off-the-shelf app does 90 percent of what you need, buy the app. Jobber, QuickBooks, and a Calendly link will run a small service business a long way before custom software earns its keep. Custom is worth it when the tool becomes your actual advantage, or when the off-the-shelf options force your business to bend around their assumptions instead of the reverse.
The honest test is volume and pain. If a task happens dozens of times a week, follows real rules, and someone dreads it, automating it returns money fast. If it happens twice a month and takes ten minutes, leave it manual and spend the budget somewhere it hurts. We turn work down on exactly this basis: a $9,000 build to save two hours a month is a bad trade, and saying so is cheaper for everyone than finding out later.
What The Money Buys When The Fit Is Right
When the volume and the pain line up, the returns aren't subtle. AltaPro built Zebra Landscaping a quoting tool that took their estimates from about four hours down to under 20 minutes. For an Alberta contractor, we built a system that cut bid assembly from two or three hours to minutes. Same pattern both times: a repetitive, high-stakes task that ran constantly, turned into something that runs in the background. That same engine is what's behind Bid Pro's, our government-bid and AI proposal platform launching soon.
None of that started with a big number. It started with one expensive task, done often, worth fixing. Find yours first, put a real dollar figure on the pain, and only then talk to anyone about a build.