Last month I wrote about the 73% of Canadian small and medium businesses that haven't thought about AI yet. A few people reached out to say the number surprised them. A few others said it didn't.

Google Cloud surveyed over 3,400 senior business leaders this year and found that 74% of companies using AI are already seeing a return. Among the earliest movers, that number hits 88%. That gap, between the average adopter and the early adopter, is the part worth sitting with.

What "seeing a return" actually means

Survey numbers like this get thrown around without much context, so it's worth breaking down what they cover. The gains show up in productivity first: 70% of organizations reported meaningful improvement. Customer experience came in at 63%. Straight business growth, revenue or margin, was 56%.

The productivity number is the one that matters most for a trades or manufacturing business, because it's the most direct. Among organizations reporting productivity gains, 39% said their employees were doing the same work in roughly half the time. Not double the output for the same hours, though some businesses got that too, but the same output for half the hours.

For a company running on tight margins, that's the whole game. If your estimator gets quotes out twice as fast, you're bidding on more jobs without hiring anyone. If your office admin stops spending Tuesday mornings chasing down PO numbers, that time doesn't vanish. It goes somewhere. Usually it goes to the next fire, the next client call, the next thing that was getting ignored because there weren't enough hours in the day.

The compounding effect nobody mentions

Here's the part the survey doesn't say outright, but the math implies it: early movers aren't just ahead, they're pulling away. Say a construction company adopts AI-assisted estimating in January and gets its quote turnaround from three days down to same-day. Every month that passes, that company is bidding more jobs than a competitor still quoting the old way. By the time the competitor gets around to it, in June, say, the early mover has six months of extra bids, extra wins, and extra cash flow behind them. Plus six months of knowing what works in their specific business and what doesn't.

That second part is easy to undervalue. The first attempt at using AI for a real workflow is usually rough. You learn what your team actually needs versus what the vendor demo promised. You figure out which tasks the tool is genuinely good at and which ones still need a person checking the output line by line. Six months of that trial and error is worth more than the six months themselves, because it's the difference between an AI tool sitting half-used and one that's actually built into how the business runs.

None of this is unique to AI. Early movers in any operational shift tend to get further ahead than the raw numbers suggest, because capability compounds. What's specific to this moment is how fast the gap is opening, and how many Okanagan businesses are watching it open from the sidelines.

Why construction sits near the bottom

Construction is still near the bottom in BC for AI adoption. An industry that figured out how to build a 40-storey tower is still deciding whether to try software that organizes documents. There's no punchline, that's just where things stand.

I don't think this is because people in construction are behind on technology generally. Plenty of the GCs and subs I've worked with run tighter operations, on paper anyway, than a lot of software companies I've dealt with. The hesitation comes from somewhere more reasonable: construction runs on trust built over decades, on relationships with subs and suppliers and inspectors, and on processes refined through actual failure. Nobody wants to hand a piece of that over to a tool that might get it wrong on a job where getting it wrong costs real money.

That instinct isn't wrong. It's just aimed at the wrong target most of the time. The AI tools that actually help a trades business aren't the ones making judgment calls on site. They're the ones handling the paperwork nobody wants to do anyway: drafting the first pass of a proposal, summarizing a stack of RFIs, flagging when a schedule change affects three other trades. Nobody's trust in their sub relationships is at stake when AI drafts an email that a human reviews before it goes out.

A common objection worth answering directly

The pushback I hear most is some version of "we tried a tool and it didn't work." Usually what happened is the business bought a generic tool built for a different industry, pointed it at their business with no setup, and expected it to understand their processes out of the box. Generic tools guess from industry averages. A tool with no idea what your actual job costs, your actual client history, or your actual crew capacity look like is going to produce generic, occasionally wrong output. That's not proof AI doesn't work in construction. It's proof that a tool with no context about your business won't do much for your business.

"We're too small for this" doesn't hold up anymore

The most common reason I hear for waiting is size. "We're twelve people, we don't have an IT department, this is for bigger companies." I understand where that comes from. Most AI marketing pictures enterprise software teams with dedicated engineers, not a five-person electrical contractor in Vernon.

But the tools that actually move the needle for a small operation don't need an IT department. They need someone willing to spend a few hours setting them up around one specific task, and someone willing to check the output for the first few weeks until it's earned some trust. A twelve-person shop doesn't need a chief AI officer. It needs one person who owns the decision to try something, plus the discipline to measure whether it worked.

If anything, smaller businesses have an advantage here that gets overlooked. There's no committee to convince, no six layers of approval, no legacy software locking you into a vendor relationship from 2011. The owner decides, the team tries it, and the results show up within weeks instead of quarters. The businesses I've seen move fastest on this aren't the biggest ones. They're the ones where the person who owns the problem is also the person who can say yes to fixing it.

What a manufacturing floor version of this looks like

Construction gets most of the attention in these conversations because industry-wide adoption numbers get tracked and published. Manufacturing faces a version of the same gap, it's just less visible from the outside.

Say a small parts manufacturer in Penticton runs three shifts and tracks production against orders using a legacy ERP system alongside a shared spreadsheet nobody fully trusts. Every week, someone spends half a day reconciling what the ERP says shipped against what the spreadsheet says was ordered, chasing down the gaps by phone and email. That reconciliation work isn't complicated. It's tedious, error-prone, and exactly the kind of pattern-matching task AI handles well once it's connected to both systems.

Get that right and the half-day becomes twenty minutes, and the twenty minutes catches more discrepancies than the half-day did, because it isn't fighting fatigue by hour three. That's not a transformation. It's a Tuesday afternoon getting shorter. Multiply that across every recurring reconciliation task on a shop floor and the hours add up fast, week after week, which is where the early-mover gap in the Google Cloud numbers actually comes from in practice.

The trust problem underneath the adoption gap

Most business owners I talk to have a trust question somewhere in the conversation, and it's a fair one. A PwC Canada survey of 220 Canadian organizations looked at this directly. A third of Canadian businesses have no plan for responsible AI use. 65% say the core problem is that nobody has been formally put in charge of it. Unclear ownership is the bottleneck, and that's a management problem, not a technology one.

This shows up in a specific, predictable way inside a trades or manufacturing business. Someone on the team starts using an AI tool on their own, maybe to draft emails or summarize documents, without anyone deciding whether that's okay, what data it's allowed to touch, or who checks the output. It works fine until it doesn't: a client's project details end up somewhere they shouldn't, or an estimate goes out with an AI-generated number nobody double-checked. Then leadership either bans the tools outright or ignores the problem and hopes nothing goes wrong. Neither of those is a plan.

71% of Canadian organizations expect positive financial returns from investing in AI governance. That number tells you something: this isn't a compliance exercise people tolerate because they have to. Businesses that build the guardrails, who can use what, on what data, checked by whom, before scaling up tend to move faster afterward, because the questions that would otherwise stall a rollout have already been answered. The companies skipping this step are the ones you read about later for the wrong reasons.

The businesses pulling ahead aren't the ones using the most AI. They're the ones who know exactly which two or three things it's doing for them, and why.

What "having a plan" actually requires for a small business

The PwC numbers make governance sound like a corporate initiative with a steering committee and a slide deck. For a fifteen-person contracting company, it's much smaller than that, and it doesn't need to be complicated to count as a plan.

It means deciding, on purpose, which tools people are allowed to use and for what. It means deciding that client financial details don't go into a free AI chatbot with no data agreement, but that a properly configured tool with the right protections can handle a draft proposal. It means naming one person, could be the owner, could be the office manager, who's responsible for knowing what's being used and checking in on it now and then. That's the whole plan: three decisions, one person accountable, written down somewhere everyone can see it.

Most businesses skip this not because it's hard, but because nobody's asked them to do it yet. Nothing has gone wrong. That's precisely the moment to put it in place, before something does, rather than after, when the fix costs a lot more than the plan would have.

What moving early actually looks like

The businesses I work with that are seeing results didn't start with an ambitious company-wide AI strategy. They started the same way, every time: one problem costing them real time or money, one specific thing to try against it, and a measure of what actually changed.

Take proposal writing. A five-person contracting shop might spend six or eight hours a week on quotes: pulling numbers from past jobs, writing up scope, formatting the document, sending it out, then doing it all again for the next lead. That's not skilled work exactly, it's assembly work that happens to require someone skilled to do it because nobody's built a shortcut. An AI tool trained on that company's own past quotes can produce a solid first draft in minutes. The estimator still checks the numbers and adjusts scope. But six hours becomes ninety minutes, and that difference shows up in how many bids go out in a month.

Or take client updates. A project manager sending five or six "here's where we're at" emails a day isn't doing anything that requires judgment. It requires time. AI can draft those updates from the day's site notes, and the PM reviews and sends. The judgment stays with the person. The typing doesn't.

Neither of these examples is complicated, and that's the point. The businesses gaining ground aren't running some sophisticated AI operation. They picked one real cost center, tried something narrow against it, and measured whether it actually helped. The clarity came from doing that, not from reading about it.

The window is still open

The gap between businesses using AI and those still watching is widening, and it's going to keep widening for a while yet. Moving now still puts you ahead of most Okanagan businesses in your industry. That won't stay true forever. The window is still open, but it's not getting cheaper to climb through.

If you want a clearer picture of where AI would actually help in your business, rather than a generic pitch, the free assessment is built to answer exactly that in about twenty minutes, before you spend a dollar on anything.