How to Build a Drone Force: Ukraine's Playbook

A drone force is four things at once, and procurement systems are built to acquire only one of them. The equipment is the visible part and the least durable: airframes, controllers, payloads and ground stations that a contested environment will make obsolete faster than a procurement cycle can replace them. The other three are people, training infrastructure and doctrine, and none of them arrive in a crate.

• Operators. Pilots who can fly under jamming, at night, against a moving target, with degraded video and a hostile electromagnetic picture.
• Instructors. The multiplier. A force with instructors produces operators indefinitely; a force with operators produces a cohort and then stops.
• Training infrastructure. Simulators, ranges, spares and a maintenance line, sized to the throughput the force needs rather than to the size of the first cohort.
• Doctrine. Unit structure, tasking, communications, resupply, rotation, and the relationship between unmanned units and everything else on the battlefield.

The reason this know-how sits in Ukraine is volume. Output ran at roughly 800,000 drones in 2023, 2.2 million in 2024 and at least 4 million in 2025, with a stated 2026 plan of more than seven million announced by the deputy defence minister at NATO’s Operational Force Development Framework conference in January 2026. That scale forced the creation of everything around the equipment: selection, training throughput, sustainment, and a doctrine cycle measured in weeks. The evidence for what allied forces have already absorbed from it, and where observation stops working, is set out in what NATO militaries are learning from Ukraine’s drone war.

Programmes fail in a recognisable order. They buy equipment first, discover they have nobody to fly it, run a course, and end up with a cohort that cannot be replaced and a doctrine gap that nobody owns. The inverse sequence costs less and stands up faster.

1. 1Define the capability target, not the shopping list. What effect the force needs to produce, at what range, under what electromagnetic conditions, at what tempo. Everything downstream — platform choice, cohort size, training throughput — falls out of this answer.
2. 2Select the first cohort for aptitude.Flying skill under stress is not evenly distributed and does not correlate neatly with rank or existing aircrew qualification. Selection discipline at this stage is what keeps the pipeline’s output high later.
3. 3Stand up the training system before the fleet.Simulators adapted to the force’s controllers and platforms, a range that allows failure, and consumables budgeted for the crash rate that learning actually produces.
4. 4Train the trainers in parallel, not afterwards. The instructor cadre should come out of the first cohorts, be certified against a written standard, and take over delivery while the advisers are still present to correct them.
5. 5Write the doctrine as the force learns it.Tactics, techniques and procedures captured while the cohort is training, reviewed against combat experience, and issued as the force’s own documents rather than a translated foreign manual.
6. 6Hand over and hold a review cycle. A programme that cannot be run without its advisers is a dependency. The handover milestone is the point of the whole exercise, and the review cycle is what keeps the doctrine current after it.

The distinction that matters in an instructor is not flight hours; it is whether they have flown against an adversary that adapts. Ukrainian government figures reported in early April 2026 attributed 33,988 of just over 35,000 Russian losses in March to drones, with artillery and other weapons accounting for 1,363 — figures published by the Ukrainian side and worth treating as such, but indicative of how central unmanned systems have become to the fight and how much practical experience sits behind the number. Western officials had put the drone share of battlefield casualties at around 80 percent a year earlier.

An instructor cadre built from that experience teaches things a syllabus does not contain: what a degrading video link means about the jammer’s position, when to abort, how to fly a profile that survives the last three hundred metres, how a crew recovers after losing three aircraft in an hour. The transfer works because those judgements can be taught deliberately once someone has made them under pressure, and it fails when the programme substitutes a manual for the person.

Simulation is where most of the training hours live, because the alternative is airframes and a crash rate. The requirement is easy to state and frequently missed: the simulator has to be connected to the controllers and platforms the force will actually field. Stick response, latency, video behaviour and failure modes differ enough between systems that training on a generic core builds habits the force pays to unlearn.

The practical route is a battle-tested training core adapted to the receiving force’s hardware, rather than a bespoke build or an off-the-shelf licence. Adaptation is a scoped engineering task with an owner and a date; it belongs in the programme plan alongside the cohort schedule, because a slip there stalls everything behind it.

Equipment and operators produce effect only through the command layer, and this is the part that rarely appears in a capability budget. Four questions decide whether a drone force works.

Allied institutions are building their own machinery for this — a joint analysis and training centre with Ukraine, and a layered counter-UAS experimentation campaign among them. Those give a force a place to compare notes. They do not, by themselves, produce operators.

Equipment selection runs alongside the pipeline rather than ahead of it, and it has three steps that are worth separating.

1. 1Evaluate against your requirement.Combat use in Ukraine tells you a system survived a real environment; it does not tell you it fits your force’s frequencies, climate, logistics or rules of engagement. What battlefield use does and does not establish is set out in what battlefield testing actually proves.
2. 2Decide between buying units and licensing production. Buying units is faster and keeps the supplier responsible for iteration; licensed or joint production builds domestic industry and takes longer to reach first delivery. The buying route in detail is the procurement mechanism; a technology licence is a separate transaction with its own permissions.
3. 3Field with the pipeline behind it. Equipment fielded to units without trained crews, spares and a maintenance line produces holdings rather than capability, and the gap shows up in the first exercise that stresses it.

Capability transfer is an export-controlled activity. Technical data and technical assistance — drawings, specifications, software, manuals, instruction and training — are controlled goods in Ukrainian law, so what may be taught, to whom, and in what form is settled through permissions before it is settled commercially. For the receiving force this is a scheduling fact rather than an obstacle: permissions have processing times, and a programme plan that ignores them slips.

• Scope under NDA first. Force size, platforms, timeline and classification level shape what can be delivered and where; that conversation precedes any documentation moving.
• Match the classification.Delivery on the receiving force’s ground, to its own standards, is often the only way to run the parts of a programme that cannot leave its system.
• Document the transfer. What transferred, to whom and under which permission is a record both sides need later, particularly where onward use or third-party participation is contemplated.

Timelines vary with force size and starting point, but the shape is consistent: an early scoping and selection phase, immersion for the first cohort and the future instructor cadre, simulator adaptation running in parallel, embedded delivery on the force’s own base, then certification of the cadre and handover with a standing review cycle. The pacing item is almost never the equipment. It is instructor availability, permissions and the simulator adaptation.

The funding environment reflects how widely this is now understood. The Drone Coalition, proposed by Latvia and co-coordinated with the United Kingdom, was formally established in February 2024 and has grown to twenty member states; it allocated about €2.75 billion for 2025 and €4.5 billion across 2024 and 2025, with roughly €180 million contributed to a UK-led joint procurement fund. Money for drones is not the constraint any allied force reports. The constraint is the force behind them.

Where a force wants the capability without standing up production, the neighbouring routes are purchase and validation rather than transfer: buying combat-proven systems is covered in the procurement guide, and the full set of ways into this market — including which of them fits a ministry with a capability mandate — is mapped in the six ways into Ukraine’s defense market.