EP 79 — PrimerAI’s Sean Moriarty on Time-Bounded Mandates vs Incremental Reform in AI Adoption

Primer.ai reduced a 50-day intelligence analysis timeline to hours, enabling operators to redirect friendlies away from hostile forces before collision. CEO Sean Moriarty explains why this speed advantage stems from working backward from customer need rather than leading with AI capabilities — understanding workflow bottlenecks where analysts spend more time assembling information than analyzing it, then building software that starts their day where it used to end.   Sean's product background shaped his counter-intuitive approach to government AI deployment: aggressive, rapid vetting with proper oversight beats trying to solve hypothetical risks in abstract policy frameworks. He also advocates for time-bounded capability mandates that force elimination of bureaucratic barriers rather than incremental process reform. Sean's team matches customer shift work during crises, a level of commitment he attributes to hiring people who care about mission as deeply as the end users they support.    Topics discussed:   Transitioning from commercial tech leadership to applying AI for national security intelligence analysis and defense missions. Building products by working backward from customer workflow bottlenecks rather than deploying AI capabilities just for the sake of it. Operating at the app layer with workflow tooling beyond LLMs, including source ingestion and closed-loop reporting systems. Mitigating LLM hallucinations through generative chunk playback against source material and surfacing contested claims with full source transparency. Deploying capabilities via API to augment existing platforms rather than forcing workflow disruption through full system replacement requirements. Accelerating government technology adoption through time-bounded capability mandates that eliminate bureaucratic barriers versus incremental process reform approaches. Reducing 50-day OSINT analysis timelines to hours by parsing massive unstructured data volumes for signal extraction and collision avoidance.