Image credit: NATO. Source: "NATO's DIANA selects innovators to drive decision superiority for the Alliance", NATO, 13 July 2026.
NATO's Defence Innovation Accelerator for the North Atlantic (DIANA), working with Allied Command Operations (ACO), has selected Flai as one of ten innovators for its Decision Superiority for NATO Warfighters challenge.
The challenge targets a specific operational problem: how modelling and simulation, targeting support, and operational wargaming can be made faster and more adaptive for the people planning and running operations. Selected companies develop their solutions and will demonstrate them to NATO stakeholders in December 2026.
Our contribution is the layer most decision-support systems assume already exists: analysis-ready 3D terrain, produced automatically from raw sensor data.
What the challenge is asking for
DIANA and ACO structured this challenge around Maven Smart System NATO (MSS NATO), an AI-enabled platform that consolidates data from multiple sources to support situational awareness, targeting, and operational planning.
The task is to augment it - to show how AI, machine learning, and software tools can extend what ACO can do with the data it already holds, and support more adaptive, real-time decision-making.
The ten selected innovators are Matrix Pro Sim, Hadean, Watchtower Labs, Flai, Grist Mill Exchange, Decent Cybersecurity, ETE Technology, Onebrief, Picogrid, and Levato AS.
Selection came through DIANA's challenge process, run in close cooperation with ACO, which means the requirements were written by the operational side rather than inferred from a market study. Development work happens with NATO stakeholders, including operational end users, in the loop.
What Flai brings to MSS NATO: 3D spatial intelligence
Decision-support systems are only as good as the spatial data underneath them. And spatial data arrives in a form nobody can act on.
A LiDAR survey, a bathymetric run, or a photogrammetric flight produces tens or hundreds of millions of unclassified points. Every point has coordinates. None of them carry meaning. Nothing in that file distinguishes bare earth from a building, a treeline from a wire, or a navigable channel from a shoal.
Turning that into something a planner or a targeting cell can query means classification: assigning each point a semantic class, then extracting the terrain models, obstruction maps, and 3D features that downstream systems consume.
Done manually, that step takes an analyst hours to days per dataset. In an operational context, that is often the difference between data that informs a decision and data that arrives after it.
Flai automates that step. Our models classify point clouds from LiDAR, bathymetry, and photogrammetry into analysis-ready 3D spatial information - currently 29 classes across 7 domain-specific pre-trained models, with F1 above 0.98 on key classes such as ground. Classification runs at under 3 minutes per km² at 16 pts/m². In production mapping workflows, teams report 30-80% less classification time and up to 50% less QC time on urban datasets.
Those numbers come from civil production work: national mapping campaigns, utility corridors, forestry inventories. The underlying problem in a defense scenario is the same one, under tighter time pressure and stricter constraints on where the data may go.
Deployment is the other half of the requirement. Flai runs where the data already is - on your own hardware, self-hosted behind your firewall, or air-gapped at the edge - so nothing has to leave the network to be classified. Our defence and security page covers the deployment options in full.
"The models we are bringing to NATO are the same ones running in civil production today, across more than 200 organisations doing national mapping campaigns, utility corridors and forest inventories. That volume is why I can say the automation holds up, instead of hoping it will. What this challenge adds is proximity to the decision. Selection is a starting position, and December is where it gets checked."
Luka Rojs, CEO and co-founder, Flai
Working with Flai on defense and security programmes
Selection into NATO DIANA is recognition that automated 3D terrain intelligence belongs in the decision chain, and an opportunity to prove it with the people who would use it.
If you are working on a defense, security, or national mapping programme with sovereignty or air-gap constraints, get in touch. We can walk through deployment options and run your own data through the models.