The AI isn't the part that makes you productive

Article

Somebody at your company signed up for an AI tool this year. Maybe it was you. A few months in, the honest assessment usually sounds like "it's fine." Quotes get drafted faster. Emails write themselves. Nobody can point to the line on the P&L where it showed up.

That's not a failure of nerve, and it probably isn't the wrong vendor. Statistics Canada has now measured the thing directly, and the finding is uncomfortable enough to be useful.

The premium was real until they controlled for the boring stuff

In April, Statistics Canada published a study linking multiple waves of its digital technology survey to administrative business data — real firms, real tax records, not a vendor poll. The headline number looks like the pitch deck you've been shown: AI adopters had 16.8% higher labour productivity than non-adopters.

Then the researchers started adjusting. First they accounted for how productive each firm was before it adopted AI. The premium dropped to 10.2%. Then they added what the study calls complementary capabilities — R&D spending, cloud computing, data analytics, robotics, and IT training for staff. The premium fell to 5.1% and stopped being statistically significant.

In plain terms: firms using AI are more productive, but not because of the AI. They were already ahead, and they'd already built the unglamorous machinery that makes any software worth owning. The authors put it about as bluntly as a government agency can: AI adoption on its own "is likely insufficient to deliver transformative productivity gains."

Data analytics is the real tell

The same study looked at which firms adopt AI in the first place, and one factor dwarfed the rest. Businesses already using data analytics were 15.0 percentage points more likely to adopt AI. Robotics came in at 8.1 points, IT training at 3.0, cloud at 2.8, R&D at 2.0.

"Data analytics" sounds like it means hiring someone with a statistics degree. In a thirty-truck fleet or a four-location retail operation it means something much less exotic: your numbers live somewhere a computer can read them, and a real person looks at them on a schedule.

Plenty of good businesses fail that test. Dispatch is in one system. Fuel cards are in another. Maintenance is in a binder. Margin by lane, or by store, or by product line, is in the owner's head — and the owner is genuinely good at it, which is exactly why nobody has written it down in twenty years. Put an AI layer on top of that and you get confident-sounding output built on nothing. The common advice is to adopt AI and let it force you to clean up your data. That's backwards. The cleanup is the investment. The AI is what you buy afterward.

Almost everyone has the tools. Almost nobody has them connected

CFIB's national survey work makes the same point from a different direction. 92% of Canadian small businesses use digital tools , but only 10% have fully integrated them across operations. The businesses that did integrate report $2.40 back for every $1 invested, against $1.60 for everyone else.

So the divide isn't between businesses with technology and businesses without. Nearly everyone has technology. The divide is between businesses where the systems talk to each other and businesses where the same invoice gets typed in three times by three people who each think the other two are the ones doing it properly.

Four questions worth sitting down with before you approve another software subscription:

Where does the same number get entered more than once? That's your first integration, and it usually pays for itself on labour alone.

Can you pull twelve months of revenue by customer, by product, or by route without asking anyone to build it? If not, that's the gap the study is describing.

Who owns the numbers when the person who knows them is on vacation?

Which single process eats the most hours and produces the least judgment? Start there — that's where a tool has something to actually do.

The tool and the training aren't separate purchases

CFIB found that Canadian firms using generative AI save an average of 1.08 hours per day . That's a real number, and it's the one worth chasing. But it shows up for people who know how to use the thing, which is why firms investing in AI turn out to be 5.4 percentage points more likely to also invest in employee training.

That same CFIB research found AI is the most polarized line in the 2026 budget: 27% of businesses plan to spend more, and 52% have no plans to spend at all. If you're in the 52%, the useful move this year probably isn't a subscription. It's connecting two systems and teaching one person to read the output.

What to take from this

Nobody is arguing AI doesn't work. The argument is about sequence. The productivity gains show up in firms that had their data in order, their systems connected, and their people trained — and those three things are worth doing whether or not you ever buy an AI tool. Punjabi-Canadian owners are heavily concentrated in trucking, construction, retail and agri-food, which are precisely the sectors where the operational knowledge is strongest and the record-keeping is loosest. That combination is an advantage waiting to be written down, not a weakness.

If you're weighing this and want to compare notes with people running businesses your size, come to one of our monthly socials — the conversation about what actually worked tends to be more honest in a room than in a sales meeting.

This is general information, not professional or financial advice. Talk to a qualified advisor before making technology or capital investment decisions.

Sources

Statistics Canada, "Artificial intelligence adoption and productivity in Canadian firms," Economic and Social Reports, April 22, 2026

CFIB, "Digital adoption including AI paying off for SMEs, but gaps remain," September 29, 2025

CFIB, "AI Adoption and Workforce Training Investment in Canada: Driver or Deterrent?" April 2, 2026