Yes, this newsletter contains AI news. 

Please don’t close the tab juuuust yet.

I know lots of AI news out there can turn eyeballs into the back of one’s skull.

Each of the three stories below points to something real happening inside companies right now, from how work is getting reorganized to who controls the infrastructure the rest of us are building on.

I pulled three recent stories that actually matter if you run a business, build products, or make decisions. 

For each one, I’ll translate the business reality and give you a question you can bring into your next meeting.

And yes, we’ll have a little fun with it too.

Too many conversations about AI still focus on layoffs. That’s the easiest storyline to understand. Inside companies actually deploying these systems, something more structural is happening. Workflows are shifting, authority is moving, and pieces of work are being reorganized in ways most companies didn’t anticipate.

My colleague Jennifer Friesen looked at this shift in a Digital Journal story on what she calls the AI jobs paradox.

REALITY CHECK: AI automates tasks, not whole jobs. Companies expecting the technology to eliminate entire roles usually end up disappointed. The organizations seeing real gains start by breaking work into individual tasks and figuring out which pieces machines can handle, which still need humans, and which tasks probably shouldn’t exist at all.

That exercise often exposes something awkward: Many workflows were inefficient long before AI showed up. Which is awkward if the robot just arrived and immediately noticed your process is held together with Slack threads and vibes.

New job alert: someone has to run the machine

Once AI enters a workflow, a different kind of work appears. Someone has to turn messy business problems into instructions the system can follow. Someone else has to check the output when the tool hallucinates something or confidently produces garbage.

In other words, someone now has the job of babysitting the robot.

Which sounds ridiculous until you realize half of corporate history is just people supervising other systems that occasionally break, with people who don't always get it.

Those oversight roles are becoming essential inside companies experimenting with AI. Managers are discovering this shift firsthand.

MYTH BUSTED: middle management isn’t disappearing. The job is simply changing. Less time assigning work, more time reviewing outputs, resolving edge cases, and making sure automated systems stay aligned with reality.

Another lesson companies keep learning the hard way.

FIX THE WORKFLOW BEFORE ADDING AI.

If the underlying process is inefficient, automation usually just makes the inefficiency happen faster. The organizations seeing meaningful productivity gains are redesigning the process itself instead of bolting technology onto the old one.

If your process is a mess, AI just helps you produce the mess faster. Congratulations. You’ve invented automated chaos.

MONDAY MORNING MEETING QUESTION - Ask the room:

“Which parts of our team’s work could AI actually handle today, and which parts still need a human?”

You’ll quickly realize jobs aren’t disappearing. Pieces of them are.

Read the full article by Jennifer:

Canada played a major role in the research breakthroughs behind modern artificial intelligence. Geoffrey Hinton and Yoshua Bengio helped push the field forward and trained generations of researchers who now work across the global AI ecosystem.

The industry built on that research, however, and grew somewhere else.

We've heard this before. Canada loses its own opps all the time.

I explored all this in a recent piece I wrote this week built on research from two people much smarter than me. 

THE PART MOST PEOPLE MISS: AI is an infrastructure business.

Training and running modern models requires enormous investment in chips, data centres, and energy capacity. The companies controlling that infrastructure increasingly control the market built on top of it.

The labs produced the ideas. The platforms captured the industry.

Why this sucks: It’s a bit like inventing electricity and then watching someone else build the power grid.

Most organizations experimenting with AI will not build their own infrastructure. They will build tools and products on top of someone else’s platform instead.

When this rolls out in a company, an AI strategy quickly becomes a cloud decision.

Once your workflows, data pipelines, and internal systems are tied to a particular provider, switching later becomes complicated and expensive. Read: BRUUUTAL.

MYTH BUSTED: the cloud didn’t eliminate vendor lock-in. It just changed what lock-in looks like.

Companies now depend on ecosystems of APIs, models, infrastructure, and developer tools that all connect to the same provider. Leaving that ecosystem later can mean rebuilding half your stack.

TL:DR: Lock-in just got a makeover. Now it comes with APIs, dashboards, and a very friendly developer portal.

TRY THIS IN YOUR NEXT LEADERSHIP MEETING 

“Which company’s AI ecosystem are we actually building on… and how hard would it be to move if we needed to?” 

Most organizations never ask that question until they’re already locked in.

Here's what all this means for Canada...

Digital sovereignty used to sound like something discussed in policy papers and government strategy sessions. Still is... but now it’s showing up in procurement meetings and boardroom conversations.

The reason is simple.

Where your data lives determines which laws apply to it and which governments can access it.

For years companies talked about “the cloud” like it was a magical place floating somewhere above the earth.

Turns out it’s still a building full of computers in a specific country with specific laws.

REALITY CHECK: the cloud still has a physical location. Infrastructure decisions that once looked purely technical now carry legal and geopolitical implications. Companies are starting to evaluate cloud providers the same way they evaluate supply chains.

Reliability matters.
Cost matters.
Jurisdiction now matters too.

AI adds another layer to the equation.

Many AI systems are trained on proprietary company data. Once those systems become embedded in daily operations, that data becomes strategically important.

Control the data and you control the system.

Give it away and the balance of power shifts.

UNCOMFORTABLE DISCOVERY —

When companies map their digital infrastructure, many realize they don’t actually control their full stack. Critical systems often run across multiple vendors, multiple clouds, and multiple software layers.

When something breaks, responsibility can become surprisingly difficult to untangle.

The implications of that shift are explored in another by Jennifer this week.

Most organizations discover this the same way they discover where all their software subscriptions live — during a mildly panicked internal audit.

QUESTION TO DROP INTO A TECH OR RISK MEETING

“Do we know exactly where our most important company data lives and which jurisdiction governs it?”

If nobody can answer that quickly, that’s the real issue. And you should be afraid.

If the answer is “uh… I think it’s in the cloud somewhere,” congratulations.
You’ve just found the actual problem.

Read the full story:

At first glance these stories look unrelated. One is about jobs, one about Canada’s role in AI, and one about digital sovereignty.

They’re actually about the same thing.

AI isn’t just introducing new tools. It’s reshaping how work gets organized, who controls the infrastructure companies depend on, and where the real leverage sits in the technology stack.

Inside organizations, the change shows up in workflows and decision-making. At the industry level, it shows up in who owns the platforms everyone else builds on. At the geopolitical level, it shows up in where data lives and who governs it.

Different angles. Same shift.

Which means the real question for most companies isn’t “should we use AI.”

It’s how much control they want over the systems they’re about to depend on.

Until next time, thanks for laughing and crying along with me,

— Chris Hogg