Why accelerating decision-making—not just building models—is the real frontier of defense AI
Guest: Mack Ohlinger – CEO, Dunedain
Most AI conversations in defense focus on models, data, or compute. This one does not. Mack Ohlinger, CEO of Dunedain, is building agent-based systems designed to compress military planning and decision-making timelines from days to seconds. This episode breaks down what actually matters: architecture, user interaction, and the hard reality of getting AI from demo to production.
Topics
Building agent-based AI systems for military planning workflows (MDMP/MCPP)
Why most AI companies fail between demo and production
The role of architecture, testing, and user interaction in mission-grade AI
Takeaways
The core constraint in defense is decision speed, not access to data or models.
AI systems that succeed mirror existing cognitive workflows and compress them, rather than replacing them outright.
The gap between demo and production is driven by integration, edge cases, and real user behavior—not model performance.
High-quality data is necessary but insufficient; systems must continuously learn from users in real environments.
Modular, MOSA-aligned architectures introduce integration complexity that must be actively managed.
Timestamped Highlights
[00:00] – Intro and mission: accelerating decision-making in defense
[02:30] – From operator experience to startup: identifying the planning bottleneck
[07:00] – Why most AI demos fail to survive contact with real users
[10:30] – The challenge of testing stochastic, multi-agent systems
[18:00] – How to frame AI capability for defense customers (KPPs, outcomes, trust)
[26:00] – Decision-making under changing requirements and dynamic missions
[28:30] – The future: AI-enabled planning from the CoCom level to the individual operator
Podden och tillhörande omslagsbild på den här sidan tillhör
Callye Keen. Innehållet i podden är skapat av Callye Keen och inte av,
eller tillsammans med, Poddtoppen.
