Next-Gen Army C2 Tests: What AI Changes in Battle

The Army’s next-generation command and control (C2) prototype is going back to the field for a second test only three months after its July contract award. That pace is the story.

When a modern force talks about “faster decision-making,” it’s easy to picture a better map or a cleaner dashboard. The reality is more operational—and more political. Speed in C2 isn’t just a UI problem; it’s a procurement model, a data governance problem, a cybersecurity posture, and a leadership habit. If any one of those lags, commanders still wait.

This post is part of our AI in Defense & National Security series, and it focuses on what this prototype effort signals: the Army is treating AI-enabled C2 less like a monolithic program of record and more like a continuously improving product. I think that’s the only approach that has a chance of keeping up with both the threat environment and the commercial tech cycle.

The real point of “next-gen C2” is time-to-fires

Next-gen C2 matters because it compresses the timeline from sensing to deciding to acting—especially for fires and airspace coordination. If you can shorten deconfliction, targeting approvals, and dissemination of the plan, you don’t just move faster; you reduce fratricide risk, conserve munitions, and keep formations harder to predict.

The Army’s upcoming field event (a second “Ivy Sting” iteration) is explicitly aimed at scenarios like deconflicting airspace before firing and executing faster fires workflows. That’s a practical, soldier-facing test: it asks whether the system can keep up with the tempo of real operations rather than whether it demos well in a briefing.

Why AI shows up first in C2 workflows

AI in command and control isn’t primarily about letting an algorithm “command.” It’s about scaling cognition. In a contested environment, staffs get buried under:

• Too many feeds (ISR, UAS, EW, SIGINT, cyber, logistics)
• Too many constraints (airspace control measures, ROE, no-fire areas)
• Too many synchronization requirements (fires, maneuver, medical, sustainment)

AI helps when it:

1. Prioritizes what humans should look at next
2. Summarizes what changed since the last update
3. Flags conflicts (airspace, timing, routes, fires control measures)
4. Recommends options with traceable assumptions

If you’re building for those four outcomes, you’re building AI that commanders actually use.

Prototyping in the field is also a procurement strategy

The Army isn’t only testing software; it’s testing a development method—short cycles, field feedback, and rapid iteration. This is the same philosophy behind sending “good enough to learn from” tech to units, then refining it based on soldier input.

The prototype approach pushes against a long-standing pattern in defense acquisition:

• Spend years specifying requirements up front
• Lock vendors early
• Deliver a “finished” system late
• Realize the environment changed
• Start over

The alternative is closer to how high-performing software organizations operate: frequent releases, clear user stories, and constant improvement. The Army is now trying to run field events aligned to something like a software sprint cadence—plan, build for a few weeks, test with soldiers, then immediately iterate.

Here’s my take: if the Army wants AI-enabled mission command, it has to accept “always shipping.” AI models, data pipelines, and threat tactics change too quickly for five-year refresh cycles.

What “modular partners” signals about the future C2 stack

One of the more meaningful details is the focus on bringing in multiple commercial partners under a single operational umbrella. Instead of betting everything on one prime, the prototype is integrating components from several companies—examples include tools for logistics awareness and AI integration.

That design choice implies a future C2 stack that looks like:

• A common data layer (with strict access controls)
• A set of mission apps/services that can be swapped
• Integration standards and test harnesses
• Continuous onboarding/offboarding of capabilities

In other words: competition doesn’t end at contract award; it becomes ongoing. That’s healthier. It also terrifies organizations that are used to “owning the whole box.”

Data governance is the make-or-break issue (not the interface)

AI-enabled C2 fails when data is late, mismatched, inaccessible, or untrusted. You can have strong models and slick screens, but if the system can’t answer basic questions—Who owns this data? Who can see it? What’s the authoritative source?—commanders revert to voice nets, spreadsheets, and “what I heard five minutes ago.”

Modern C2 demands governance that is operationally realistic:

• Role-based access that works at tempo (not a ticketing system)
• Cross-domain policies that balance risk with mission need
• Data lineage so outputs can be challenged and verified
• Version control for plans and overlays so teams aren’t fighting the last update

A useful working definition for readers in this space:

AI-ready C2 data governance is the set of technical controls and human rules that make battlefield data shareable, attributable, and auditable fast enough to be tactically relevant.

That last clause—fast enough—is what most governance frameworks forget.

Cybersecurity friction is inevitable—here’s how to handle it

Fast iteration creates security tension, but “slow to ship” is also a security risk. The recent public controversy around prototype security protocols (and the Army’s response that issues had already been resolved) is a predictable byproduct of moving quickly in a heavily regulated environment.

The lesson isn’t “security doesn’t matter.” It’s the opposite: security has to move at sprint speed too. Practically, that means:

Security practices that match rapid C2 prototyping

• Pre-approved secure reference architectures for prototype environments
• Continuous ATO-like evidence collection (artifact automation, not manual binders)
• Red-team-informed backlog (security findings become sprint tasks with owners)
• Feature flags so risky capabilities can be disabled without ripping code
• Golden signals for monitoring (latency, auth failures, anomalous access patterns)

And culturally:

• Don’t escalate concerns through memos as the first move.
• Pull the right operators, security leads, and program leadership into the same room.

That “conversation-first” posture is more than etiquette; it’s how you prevent security concerns from becoming bureaucratic weapons.

What Ivy Sting-style tests should measure (if we’re serious)

Field tests of next-gen C2 should produce measurable, repeatable outcomes—not just anecdotes. If your goal is lead generation in the defense tech ecosystem, this is where credibility is earned: by speaking in metrics programs can adopt.

Here are metrics that actually map to operational value:

1. Time to publish a change (plan update to propagation across echelons)
2. Time-to-fires (sensor cue to approved engagement, by scenario)
3. Deconfliction latency (airspace conflict identified → resolved → disseminated)
4. Data freshness (median age of critical tracks, overlays, logistics status)
5. Operator workload (task time, error rates, cognitive load proxies)
6. Resilience under denial (performance with degraded comms, GPS, bandwidth)
7. Auditability (can the staff explain why the system recommended option B?)

If a prototype can show improvement on even 3–4 of these with consistency, it’s not a demo anymore—it’s a capability.

What this signals for AI in defense and national security in 2026

The Army’s next-gen C2 work is part of a broader shift: defense AI is moving from “model performance” to “mission performance.” That’s a good correction.

In 2026, the winners in AI for national security won’t be the teams with the flashiest models. They’ll be the ones who can:

• Integrate across messy data sources
• Operate securely in contested environments
• Ship updates without breaking the force
• Earn trust through transparency and repeatable testing

There’s also a strategic implication: a modular C2 ecosystem makes it harder for adversaries to predict U.S. capabilities. If apps and services can be swapped and improved continuously, the “observed” system today isn’t the same system six months from now.

That uncertainty is a deterrent multiplier.

If you’re building or buying AI-enabled C2, start here

Most organizations start with features. Start with decisions. The fastest way to waste money in AI-enabled command and control is to automate tasks that don’t change outcomes.

Here’s what works in practice:

• Map the top 10 decisions your commanders and staffs make under time pressure.
• For each decision, list:
  • Inputs required
  • Latency tolerance (seconds vs minutes vs hours)
  • Trust requirements (what must be explainable)
  • Failure modes (what happens if it’s wrong)
• Only then define:
  • Data products
  • AI services
  • Interfaces

If you want a simple mantra to align teams:

Design C2 around decisions, then engineer data to serve them.

That principle scales whether you’re a brigade staff, a program office, or an industry team.

Where next-gen C2 goes next

The next phase of this prototype effort will be shaped by two forces pulling in opposite directions: the Army’s need to move fast, and the institution’s need to manage risk. The teams that thrive will be the ones that can do both without turning either into theater.

For leaders tracking AI in defense and national security, this is the signal to watch: Can the Army keep a sprint rhythm while hardening security, tightening governance, and expanding the partner ecosystem? If yes, next-gen C2 becomes a template for other modernization programs—not just a better command post tool.

If you’re evaluating AI-enabled C2 solutions (or building components that need to plug into them), now is the time to pressure-test your assumptions about data rights, deployment models, and how your product behaves when the network is degraded. That’s where adoption is won or lost.

What would your organization have to change—technically and culturally—to ship mission software on a three-week cycle without sacrificing security?