The doctrine is shifting downward
DARPA's MOSAIC warfare concept proposes replacing exquisite, monolithic platforms with modular functional nodes that can be mixed and matched via AI-enhanced networks. Instead of a single $100M platform doing everything, dozens of cheaper nodes each contribute a capability, and the system composes mission plans from whatever is available. The 2026 DARPA RFI (DARPA-SN-26-33) requests autonomous drone warfare networks with large-scale unmanned constellations and containerized support systems for exactly this kind of distributed operation.
The January 2026 DoD AI Strategy memo mandates AI that operates 'on-board, in real time, and often without any sort of connectivity or centralized compute resources.' Admiral Paparo's April 2026 INDOPACOM posture statement calls for 'AI-enabled autonomous systems' as 'significant and affordable asymmetric advantage.' Dr. Alex Miller, the Army CTO, put it plainly in October 2025: 'Automate the mental tasks up to the decision to shoot or not shoot. Let commanders focus on human judgment.'
The direction is clear across every service and combatant command: push intelligence, decision support, and autonomous capability lower in the formation. The question is what platform enables that.
Ukraine's bottom-up innovation as a forcing function
Ukraine has deployed over 4 million drones since 2022. Novel solutions go from concept to mass adoption within months. Soldiers modify commercial drones with 3D-printed components in trenches and deploy them within 24 hours. The Ukrainian DOT-Chain system works like an online marketplace where frontline units procure equipment based on immediate operational needs.
This model is messy, unmanaged, and occasionally dangerous. It also works. A Chatham House analysis from March 2025 concluded that Ukraine's decentralized approach to military innovation produced capability faster than any traditional acquisition process could match. West Point's Modern War Institute published a piece titled 'Want Drone Dominance? Let the Squad Fail,' arguing that bottom-up experimentation, even when individual efforts fail, produces faster institutional learning than top-down programs.
Every NATO army that studied Ukraine reached the same conclusion: bottom-up innovation works, but it needs a management layer. The soldiers are going to adopt commercial hardware regardless of what acquisition commands do. The choice is whether that adoption happens with classification enforcement, device management, fleet control, and audit trails, or without them.
The US military is already moving this direction
The Army Software Factory now enables soldiers across the force to build tools inside approved Army platforms. Kessel Run is launching its Next-Generation Air Operations Center. Platform One's Puckboard tool automated scheduling at the 618th Air Operations Center. The Marine Corps has its own software factory. The Army established the 49B AI/ML officer MOS and stood up Detachment 201, the Executive Innovation Corps, recruiting officers from Meta, OpenAI, and Palantir.
In July 2025, Hegseth's Drone Dominance memo directed the Army to accelerate procurement from commercial suppliers. Almost 700 frontline units are now authorized to purchase critical systems directly from commercial vendors. The War Department issued an RFI for 300,000 drones, asking industry to produce them quickly and cheaply.
Navy Task Force 59 demonstrated the model in the 5th Fleet AOR: unmanned systems, sensors, mesh networks, and AI/ML tools operated by small teams, with commercial hardware at the core. It reached full operational capability in January 2023 and has been expanding since.
What bottom-up ISR actually requires
A squad or company-level ISR capability cannot depend on enterprise infrastructure. There is no program office at the platoon level. There is no dedicated IT staff at a combat outpost. There is no reliable SATCOM on a dismounted patrol.
The platform has to run on hardware the unit already carries or can procure commercially. It has to manage those devices with classification-aware MDM that enforces data boundaries without requiring a STIG audit team on site. It has to deliver software updates to airgapped nodes without network infrastructure. It has to fuse whatever sensors are available, cameras, RF receivers, NFC readers, acoustic pickups, into a coherent picture. It has to run AI inference locally so the capability works when comms are denied. And it has to connect all of this through a mesh that routes across whatever links survive.
EdgeLance was designed for exactly this problem set. Not as an enterprise C2 platform competing with Lattice or Maven at the division level, but as the operating layer that gives a company commander, a team lead, or a medic the same kind of AI-assisted awareness that currently requires a brigade headquarters and a SATCOM uplink.
The platform layer that makes bottom-up work
EdgeLance provides six integrated capabilities that together enable bottom-up ISR without enterprise infrastructure.
Local AI orchestration runs approved models on-device, routing inference across local hardware, base GPUs, and cloud resources based on classification policy, link availability, and compute capacity. The model catalog tracks version, approval status, and export restrictions. Mission-specific loadouts package the right models and reference material for each operation and wipe them after.
EdgeLance Mesh routes data across seven link types simultaneously, co-optimizing for throughput, latency, and battery life. Store-and-forward queues events during disconnection and syncs automatically on reconnect. The mesh health feeds directly into the P(t) mission readiness score so commanders can see the operational impact of degraded connectivity.
Classification-aware MDM turns consumer hardware into hardened tactical nodes with four classification tiers, STIG compliance evaluated continuously, stealth mode for RF suppression, NVG mode for night operations, and duress PIN for compromised operators. Fleet management delivers signed software packages to airgapped nodes via Software Courier.
Evidence-coupled AI ties every assessment to its source data so operators can verify, challenge, and override recommendations. The mission record preserves the full decision chain, from sensor input through AI assessment through operator action, for after-action review, legal review, and institutional learning.
Role-specific operator views give the team lead, medic, commander, analyst, and systems admin each a workspace tailored to their job, all drawing from the same mission truth.
Complementing the enterprise layer, not replacing it
Lattice, Maven, and other enterprise C2 platforms will continue to serve brigade and above. EdgeLance does not compete at that level. It operates below it, pushing useful capability down to small units and syncing structured events, evidence, and audit trails upward when policy and connectivity allow.
A company running EdgeLance at a forward outpost generates a structured mission record: entity tracks, threat assessments with source evidence, operator decisions, mesh health data, device compliance status. When that company has backhaul to the battalion TOC, those records flow into whatever enterprise system the battalion uses. The data format is designed for interoperability, including Cursor-on-Target publishing for TAK/ATAK integration.
This is the architectural pattern that MOSAIC warfare describes: modular nodes contributing capability to a larger system, composable at the network level, independently useful when disconnected. EdgeLance is the node-level operating layer that makes that pattern work for ISR at the company level and below.