Say “dual-use technology” and most people picture hardware. Radar. Drones. Satellites. Something with a military version and a civilian version, built in a factory, with a conversion step somewhere in the middle.
That definition is out of date, and the newer one is more interesting. Dual use in 2026 does not mean a military product with a civilian spin-off, or a civilian product with a military bolt-on. It means one capability with two sets of customers and, ideally, no conversion step between them.
Canada’s clearest example runs both directions at once. The RADARSAT satellites watch ice, floods, crops and shorelines for civilian agencies. They also watch ships and coastal waters for military ones. Same capability. Two customers. Nothing is converted.
Once you hold that definition, a question follows that most software companies have never asked themselves: how much of what we build is already dual use, and we simply never noticed?
Organizing is organizing
Here is the argument, and it is simpler than it sounds.
Strip any organization down to what it actually does hour by hour, and you find the same short list. Decide what matters right now. Give it to a specific person. Confirm they have it. Track whether it happened. Adapt when it does not.
A business does this. A family does this. A platoon does this.
The domain changes the stakes enormously. It does not change the structure. The consequences of a missed handoff are wildly different in each case, but the shape of the handoff is identical: someone decided, someone was told, someone acknowledged, someone checked.
Most organizational software ignores this and builds for one domain anyway. Enterprise project tools assume a corporate hierarchy and fall apart the moment the hierarchy is a household. Consumer to-do apps assume a single user and fall apart the moment accountability matters. Both encode assumptions about the domain into primitives that did not need them.
We built InfiniteUp Tasks for households and small businesses. What we noticed afterwards is that very little of what we built is actually about households or small businesses.
Five primitives that do not care what you are
Delegation is acceptance-gated. You can assign anything to anyone, and nothing lands on their board until they accept it. We built this because people resent having tasks dumped on them. But an order issued and an order acknowledged are different states everywhere, and a system that collapses them is lying to whoever is reading the board.
Capture works without hands or eyes. Speak a sentence the way you would say it out loud, and it becomes a structured item with steps and a date. We built this for people driving home from a job site. It is the same requirement anywhere people are moving, carrying something, or wearing gloves.
Paper converts to structure. Point a camera at a whiteboard, a clipboard or a printed sheet, and the items are read back to you before anything is committed. We built this for the list on the fridge. Most organizations still running on whiteboards are not households.
Focus collapses to today. One mode that shows what is due now and nothing else. Built to reduce overwhelm. The same mechanism reduces cognitive load under any kind of pressure.
Language is not a label layer. Thirty languages, including right-to-left scripts, where the entire interface flips rather than the strings being swapped. Among them are Cree, Inuktitut and Anishinaabemowin. We built that because Canada has obligations and because our users are in 117 countries. It happens to be exactly what any coalition, multinational or community deployment requires.
None of those five were built with defence in mind. We have no military users and no defence pilot. That is rather the point. We did not design a dual-use product — we designed an organizational one, and organizational primitives do not know what domain they are in.
Why this matters in Canada, right now
The policy environment has moved faster than most founders have noticed.
British Columbia has just launched a provincewide defence and dual-use technology testbed, delivered by Innovate BC through the Integrated Marketplace program, giving companies a real-world environment to test and scale technologies with both military and civilian applications. Its definition of dual use aligns with federal and NATO standards, and the categories it names explicitly include logistics, communications and artificial intelligence.
Federally, the Defence Industrial Strategy has committed over $240 million specifically to small and mid-sized businesses advancing dual-use technologies, with the National Research Council investing over $900 million and a dedicated support stream for dual-use companies inside IRAP. The Business Development Bank raised its defence platform to $6 billion, including a $300-million fund aimed at early-stage ventures building dual-use technology.
A survey last year identified almost 500 Canadian companies already developing technologies with both defence and non-defence applications. Nearly 90 percent of them said procurement access was important to their success, and the recurring complaint was not a lack of innovation but a lack of a navigable path.
We are a four-person studio in Nanaimo. We are not a defence contractor and we are not pretending to be one. But the testbed exists precisely for companies that believe their architecture is dual use and have not yet had the chance to find out. We intend to put our hand up.
The actual thesis
Here is what I think most people are getting wrong about the AI race.
The competition is not going to be settled by whoever has the smartest model. Frontier capability is converging, it is available to anyone with an API key, and the gap between the best model and the second-best is measured in months and shrinking. Intelligence is becoming a commodity input.
What is not converging is organization. The ability to take that intelligence and actually route it — to the right person, at the right moment, with the right context, and confirm it landed — is a structural property of how an institution works. It does not arrive with a model upgrade.
We learned this the hard way inside our own studio. When we removed our AI capacity ceiling, we expected to get proportionally faster. We did not. The constraint moved immediately to our handoffs, our context loss, our human coordination. The tools were the easy part. The reorganization was the work.
Scale that observation up and it becomes an industrial argument. A country does not win on AI by having the cleverest models. It wins by being organized enough to deploy them, and by building infrastructure that serves more than one purpose at a time.
That is what double-dipping technology actually means. Not militarizing civilian tools or softening military ones, but recognizing that the organizational layer underneath both is the same layer, and that a small country facing an uncertain decade cannot afford to build it twice.
Canada has a productivity problem and a decade of catching up to do. We also have about 500 companies sitting on capabilities that work in more than one world, most of whom have never framed it that way.
The question is not whether our technology is smart enough. It is whether we are organized enough to use it twice.