Canada's AI Strategy Mistakes Adoption for Productivity

Canada has a real productivity problem. We have for decades. In 1984, the Canadian economy was 88% as productive as the US per hour worked. In 2022, it had fallen to 71%.¹

The new AI strategy is an attempt to fix that. It promises a 3 percent increase in GDP, from labour productivity. “Increase Canada’s business adoption of AI from 12 percent today to 60 percent by 2034, through boosting SMEs and business adoption supports.”² I’m reminded of the mineshaft gap in Dr Strangelove (Stanley Kubrick was satirizing the military focus on the missile gap in the 1950s). Adoption of AI doesn’t make an organization more productive. I made the same argument about the federal government: dropping AI onto a broken process just automates the mess.

Saying a new technology will magically make us more productive doesn’t make it so.

Sectors

The strategy names sectors where AI will create “outsized value”: healthcare, energy and natural resources, transportation, agriculture, and manufacturing. Why these sectors? Based on what evidence?

There are many ways AI could help and harm in all of these sectors. The strategy doesn’t say anything useful about how AI will help.

Healthcare

I will focus on the health sector, because it’s the one we hear about most often. In Ontario, the stories are of emergency rooms overflowing. In some rural communities, the emergency rooms operate on reduced hours due to staffing shortages. To make the emergency room more efficient, we might try to use AI to help in the triage process. Assuming the AI doesn’t make any mistakes (unwise assumption), we’ve sped up the triage process. Unfortunately if the hospital is operating at 100% or more of capacity in terms of the number of beds, speeding triage doesn’t help. It helps a handful leave faster but doesn’t make a meaningful difference.

Our mistake? We tackled the emergency room problem in isolation.

Putting on our Systems Thinking hat, we want to ask two questions:

• Why are there so many people visiting emergency rooms?
• Why are hospital beds so full?

Hospital beds

• Low population to hospital bed ratio
• Waiting lists for long term care
• Staffing shortages

Emergency Room Volume

• Lack of family doctor
• Lack of family doctor turns simple problems into more complex problems, which in turn might result in a hospital stay
• Unsure where to go: walk in clinic? Emergency?
• Seasonal flu et al

These factors don’t sit in neat boxes; they feed each other. Drag any node in the diagram below to see how a change ripples through the system. Notice the reinforcing loop on the left: admin burden burns doctors out, so fewer family doctors remain, which piles more admin on those still practising. That shortage pushes people into the ER, and the beds, our real bottleneck, stay full whether or not triage runs faster.

Interactive diagram made with LOOPY by Nicky Case (public domain / CC0).

None of these problems will magically get fixed by AI. If we want our healthcare system to help us get well/stay well, look at the overall system.

One example of a hospital taking a systemic approach? Humber River Hospital has a command centre where they’re paying attention³ to:

• Emergency Department volume
• Bed allocations and assignments
• Room cleaning
• Critical care capacity
• Surgical scheduling
• Discharge planning

Reading the promotional PDF from GE⁴ the command centre is focused on getting the patients who need admission a bed faster and reducing pain, complications and secondary infections.

GE promotes this system as using AI. However, the AI isn’t the central feature. The value of this system is that people in the hospital are paying constant attention to how the hospital is working. If there is a constraint, they make changes to reduce the problem. The command centre is staffed by a cross disciplinary group of people, so that there is enough knowledge in the room to make good decisions.

AI is useful here, because they’ve studied the flow of the system. They’ve established useful metrics: getting people healthy and allowing them to leave hospital. Here’s the irony: the pattern-finding that actually helps predates ChatGPT and GenAI entirely. GE brands it “AI” but it’s not the AI the strategy is betting on.

The material available in public is promotional so the actual effectiveness of the real system is something I can’t see.

AI didn’t make Humber River Hospital more effective. Improving the flow in the system made it more effective. Predictive analytics is just a tool in this picture.

Family Doctors and GenAI

Earlier we noted that the lack of family doctors is a contributing factor to over capacity.

After a full day of seeing patients, family doctors still have to write up charts documenting their findings, treatment and prescriptions. This is unpaid labour, the doctor is paid for the patient visit and the documentation is treated as overhead.

Family doctors are already dealing with burnout, it turns out that OHIP payments in many cases don’t cover the realistic costs of running a clinic (rental, front office staff, medical equipment, computers, etc.) In an attempt to reduce the unpaid labour, many doctors I know have turned to GenAI transcription tools. There are clearly documented problems with some of these tools making mistakes with both the diagnoses and medications⁵. Yet the physicians I know report saving a considerable amount of time. Possibly another example of GenAI making us feel more productive without making a big difference.

GenAI might help here, however we’ve solved a symptom. Family doctors are expected to run businesses (not their training) and are expected to handle a large number of administrative tasks along with seeing patients.

Transcription tools may help, but they don’t solve the underlying problem. The administrative burden on family doctors is a major factor in their burnout.

We need to fix the system, and not just the symptoms.

Not a Magic Wand

GenAI isn’t a magic wand and it isn’t a panacea. In the right context, it can be very useful. Used in the wrong place, it can make an organization more fragile, not more resilient.

Humber River Hospital’s command centre is a good step in the right direction. If we want to make a meaningful difference in the healthcare system, we need to look wider than just an emergency room. In short, we want to understand why family doctors are leaving the profession. If we can’t keep them happy to stay in the profession, then we can’t expect to keep patients out of the emergency room. At the other end of the spectrum, since many hospital beds are occupied by patients waiting for long term care, we need to find compassionate ways to improve the long term care system. We may also need more hospital staff and more beds.

The Healthcare system in Canada is just one example of many. We have a productivity problem, but we can’t just tell people to use GenAI and expect an improvement. If we want to improve productivity, we need to look at each organization as a system, understand the bottlenecks in the system and focus on eliminating that bottleneck.

Image attribution: Agile Pain Relief Consulting (July 2026)

Footnotes

1. Bank of Canada, Time to Break the Glass: Fixing Canada’s Productivity Problem ↩

2. Innovation, Science and Economic Development Canada, Canada’s National Artificial Intelligence Strategy ↩

3. Humber River Hospital Foundation, Humber River Hospital’s Command Centre Has Created 23 Virtual Beds ↩

4. GE Healthcare, Humber River Hospital Command Centre: One Year in Review (PDF) ↩

5. CBC News, Medical AI transcriber for Ontario doctors ‘hallucinated,’ generated errors: auditor general ↩