A product and design perspective on planning, simulation, AI, and human judgment.
The acquisition of Battle Road and the capabilities introduced through AtomEngine suggest an interesting shift: from planning as a document artifact toward planning as a living system, one where AI supports human judgment rather than displacing it.
Context
After learning more about Onebrief, Battle Road, and AtomEngine, I found myself thinking about what becomes possible when planning, simulation, AI, and human judgment exist within the same operational environment.
This is not intended as a redesign, proposal, or roadmap. It's a collection of observations, opportunity spaces, and questions that emerged while reflecting on the future of operational planning and decision support.
The intent is to think out loud, not to prescribe. Some of these ideas may already be well-understood internally. Others might be useful starting points for conversation.
The goal is not to automate decision-making.
The goal is to automate preparation so human judgment can be applied where it matters most.
Everything else on this page is an attempt to explore what that distinction makes possible, and what it demands from design.
AI can generate options. Simulation can forecast outcomes. Systems can organize information. But the most valuable decisions often emerge from human experience, intuition, creativity, and judgment. The opportunity is not replacing those capabilities. It's amplifying them.
A Shift Worth Examining
Today
Emerging Future
Hover each stage to explore. In the emerging model, each phase informs the others; the loop becomes a system rather than a sequence.
Three Observations
01
The handoff model (plan, brief, execute) assumes a relatively stable environment. What happens when conditions change faster than the plan can? When planning is document-anchored and event-driven, the plan may already be aging the moment it's published. A system-oriented approach might make planning less like a deliverable and more like a capability that's always running.
02
Plans become systems.
Not documents.
A document captures intent at a point in time. A system can represent intent across time. That difference has implications for how plans are read, shared, challenged, and updated. The design question may not be how to make a better document, but whether the document is still the right model at all.
03
Most planning tools optimize for the plan as an output. But the plan is infrastructure; the decision it enables is the real output. That reframe raises different design questions: How is uncertainty surfaced? How are hidden assumptions challenged? How does the system help a commander know what they don't know?
Human Judgment as Strategic Advantage
The value of expertise is not generating more options. It's recognizing what the system isn't accounting for.
A commander's most important contribution may not be producing another course of action. It may be identifying a hidden assumption embedded in the planning logic, an overlooked logistical constraint, a political reality the model couldn't represent, or a simpler path the AI didn't surface because it wasn't optimizing for simplicity. Great judgment often shows up as the thing that stops the process, not the thing that advances it.
AI generates possibilities. Humans contribute judgment. The strongest plans seem to emerge from the interaction between the two, not from either acting alone.
What AI can contribute
What humans contribute
The goal
Automation that creates more space for expertise, not less.
The risk in AI-assisted planning is not that it automates too little. It's that it automates so much that humans become passive consumers of recommendations rather than active contributors of judgment. Good design, I think, should expand the space of what a commander considers, not compress it.
Opportunity Spaces
Space 01
What would it mean for a plan to update as conditions change, not through manual revision, but through a continuous feed of structured intelligence? The question isn't just technical. It's a design question about what it means to trust a plan that's always in motion, and how a planner maintains their relationship with intent rather than just output.
Space 02
The ability to explore multiple courses of action simultaneously, simulated against different assumptions and compared against shared criteria, seems less like a feature and more like a different relationship with uncertainty. The design challenge is making those comparisons legible and useful rather than overwhelming, especially under time pressure.
Space 03
There may be something valuable in AI that handles analytical groundwork (drafts, assessments, comparisons) while treating human review as the essential, non-optional step. Not AI that decides. AI that prepares, so the human can contribute what only they can: the judgment about whether the preparation was asking the right questions.
Questions I'd Explore
How should AI surface hidden assumptions embedded in a plan, especially ones the planner doesn't know they're making?
When the system generates a recommended course of action, how does the interface preserve the commander's inclination to question it rather than accept it?
How should confidence be communicated without inducing overconfidence in AI-generated assessments?
How should uncertainty be visualized in a way that informs rather than overwhelms, especially under time pressure?
How do you explore multiple futures simultaneously without fragmenting a commander's mental model of the present?
Is there a version of AI staff work that actively surfaces the simpler alternative, the one that doesn't require as much coordination or risk?
How do we design for junior planners still building intuition, so that automation accelerates their development rather than substituting for it?
What does calibrated trust look like in the context of AI-generated staff work? How does design help someone know when to rely on the system and when to push back?
Why This Excites Me
There are few design problems that sit at the intersection of systems thinking, simulation, AI, and high-stakes human judgment. Most software improves productivity at the margin. This work feels like it has the potential to reshape how organizations think, not just how they document their thinking.
The simulation layer is particularly compelling to me. When a team can explore futures before committing to a path, the nature of preparation changes. It becomes less about predicting the right answer and more about building strategies that hold across uncertainty. That's a different kind of confidence than certainty.
And the design challenges here are genuinely hard. Communicating uncertainty without inducing paralysis. Preserving strategic creativity when AI is generating most of the options. Making the interaction between human expertise and AI analysis feel like collaboration rather than consumption. These aren't UI problems. They're epistemic problems with a UI surface, and I find that combination rare and worth working on.
"What excites me most is not the opportunity to build better software.
It's the opportunity to help people make better decisions."