Look, 2026 is here and the conversation in every serious room — government agencies, contractor HQs, corporate strategy sessions — has already moved past the hype. The question isn’t “Will AI change how we make decisions?” It’s “How much of this call am I willing to hand over to the model, and what happens to my judgment when I do?”
Recent data shows most executives are already using AI to support decisions on a regular basis. Some forecasts say half of business decisions will be augmented or automated by AI agents before long. In government circles the same pressure is building — faster analysis, better forecasting, tighter resource allocation. The tools are getting scary good at the “observe and orient” part of the loop.
Here’s where most people get it wrong.
They treat the machine like the new general and themselves like the staff officer who just briefs the slides.
That’s backwards. Sun Tzu never handed the army over to his best spy or his fastest messenger. He used them for what they were good at — foreknowledge, speed, volume — then he made the call. The general who forgets that distinction doesn’t get outmaneuvered by the enemy. He gets outmaneuvered by his own tools.
AI is the ultimate force multiplier for information and speed. It can scan more data, spot more patterns, and run more scenarios in an hour than a room full of analysts could in a week. That’s real power. But it still operates inside the terrain you give it. It doesn’t know which risks are politically radioactive. It doesn’t feel the weight of a decision that could end careers or put people in harm’s way. It doesn’t read the micro-expressions around the table when a recommendation lands sideways. That’s still your job.
The leaders who will actually win in this environment are the ones who keep the machine in its lane and themselves in the general’s seat.
Here’s how you do it without becoming either a Luddite or a button-pusher.
Use AI for the heavy lifting on foreknowledge — then verify like Sun Tzu’s spies.
The man obsessed over accurate intelligence because bad information is worse than no information. Modern AI gives you volume and speed the ancient world couldn’t imagine. Use it for exactly that: broad pattern recognition, anomaly detection, rapid scenario generation. But never let the output become the decision. Cross-check the critical variables the same way you always did — with human sources, ground truth, and your own read of the political and operational terrain. The machine is fast. It is not infallible, and it has no skin in the game.
Protect the terrain only humans can hold.
AI is strong where the problem is well-defined, data-rich, and relatively stable. It gets brittle fast when context shifts, values clash, or the downside is asymmetric and irreversible. That’s your ground. Ethics. Political fallout. Second- and third-order effects on people and alliances. Reputation. These are not “soft” factors — they’re the actual terrain on which most high-stakes decisions are won or lost. Let the model optimize inside the box you define. You decide where the box ends.
Read the room the old-fashioned way — even when the data looks clean.
This is where the body language work I’ve done for years becomes brutally practical. When you bring an AI-generated recommendation into a meeting, watch what happens. The slight lean back. The crossed arms that weren’t there a minute ago. The quick glance between two people who never speak up. Those signals tell you the model missed something the data didn’t capture — fear, politics, history, trust. I’ve watched too many smart teams nod along with impressive dashboards only to watch the real decision get made in the hallway afterward. The machine doesn’t see that. You have to.
When people push back on the tools themselves, redirect instead of argue.
You’re going to hear the same limiting beliefs over and over: “This is just going to replace us,” or “We can’t trust black-box recommendations on something this important.” In my book Redirecting Rejection book, I lay out the exact patterns for handling this without turning it into a fight. One that works especially well here is to acknowledge the real concern first, then reframe the tool as the thing that removes the low-value work so humans can focus on the judgment calls only we can make. The goal isn’t to win the argument. It’s to move the conversation from fear to ownership.
Build the habit of explicit human override.
Every AI-assisted process should have a clear, documented point where a named human looks at the output and says “I own this” or “We’re adjusting because the model missed X.” This isn’t bureaucracy. It’s command. It forces the discipline Sun Tzu demanded: the general stays responsible. When things go wrong — and they will — you want the after-action to land on the human who signed off, not on some opaque model. That accountability is what keeps the machine in its proper supporting role.
The uncomfortable truth is this: organizations that let AI drift into the general’s chair will look fast and modern right up until the moment a decision with real consequences goes sideways. Then everyone will be standing around asking how the model got it so wrong. The answer will be simple. It didn’t. The people who were supposed to be in charge did.
Sun Tzu didn’t win because he had better information than his enemies. Plenty of generals had spies. He won because he knew what to do with the information once he had it — and he never mistook the messenger for the man giving the orders.
AI is the most powerful messenger we’ve ever built. Treat it like one. Keep the seat at the head of the table for the person who actually has to live with the outcome.
That’s the difference between leaders who use the tools and leaders who get used by them.
The machines are getting faster every quarter. The question isn’t whether you’ll have access to better analysis. The question is whether you’ll still be the one deciding what the analysis is for.
Stay Strong, Stay Disciplined, Stay Dangerous,
-S
Full piece is also on the site here: [link to artofwarsolutions.com]
Take the free War Room Personality Codex quiz here: [quiz link]
