The Space Force's In-House AI Tools, and the Workload Surge Behind Them

The 1st Range Operations Squadron at Patrick Space Force Base runs about 90 launches a year. Two years ago, it ran 19. Headcount has not grown to match, so the squadron's guardians built the AI tools they now use to do more with the same people.

Across the smallest U.S. military service, narrow AI tools written by the operators who use them are absorbing a workload surge. A launch chatbot and a launch-communications transcriber at Patrick SFB, and a battle-management experiment in Las Vegas that compressed an hour-long tasking into minutes, are the only two with named people, named tools, and public receipts so far.

At Patrick, the 1st Range Operations Squadron built two tools through the service's Generative AI Challenge, with 2nd Lt. Hummel among the guardians leading the build. The first, called Kinetic, is a chatbot wired into the squadron's Mattermost channel that lets guardians ask mission-status questions without paging a human colleague. The second, Echo, transcribes and summarizes the radio chatter from launch communications so the same operators can scan a feed instead of re-listening to it.

The squadron's pitch is that a small team of operators can ship software that meets a specific operational need faster than a program office can. The Patrick SFB account does not publish measured labor savings from Kinetic or Echo, and the squadron's launch-count growth is best read as a workload story. Patrick SFB's launch tempo went from 19 to about 90 a year without a matching headcount increase. That is a workload surge, not a measured AI productivity result.

The second thread is at the Space Force's and Air Force's MASH battle-management experiment, run over two weeks in May 2026 in Las Vegas. MASH was the service's first attempt to integrate multiple AI subfunctions into one simulation rather than test each one in isolation. Six industry teams and 805th software engineers each plugged in three microservices: one to identify and categorize entities in the simulated battlespace, a second to match available capabilities to those entities, and a third to generate courses of action for review. An Air Force Research Laboratory orchestrator stitched the apps together so the same scenario could move from sensing to recommendation inside a single workflow.

The participant account that has traveled furthest is Capt. Adam Sochia's, a battle manager with the 552nd Operations Support Squadron. Sochia told DefenseScoop that the experiment let him complete five or six of the same battle-management taskings in the time a single one used to take him, "anywhere from 50 minutes to an hour" each.

Two limits are worth holding open. Sochia's comparison is one operator's account from one MASH simulation; the Space Force has not published a controlled benchmark, error rate, or service-wide productivity figure from the run. The Las Vegas experiment was a simulation, not a deployment, and the guardians inside it have framed it as a proof of concept, not a fielded capability. The official reporting describes AI as the advisor and the human as the decision-maker, which is what keeps the experiment a multi-tool workflow rather than an autonomous kill chain.

military.com's roundup of "six ways" the service is putting AI to work named these two categories first, with a broader inventory beyond them. The accessible reporting supports those two without completing the rest. The Patrick SFB build shows a small, named team shipping a working tool; the MASH experiment shows the service wiring multiple tools into a single workflow under simulated conditions.

The next receipt will be data analysis from MASH and, separately, whether Kinetic and Echo graduate from a single squadron's experiment to something other units adopt. The Space Force is still publishing the early returns on a workload it cannot staff its way out of.