Trust
No black boxes. Every claim footnoted. Every source open and inspectable. Every judgment traceable to the reasoning that produced it. If you can’t defend it in a room, it doesn’t ship.
Human-Led,AI-EnabledIntelligence:
An AI system for intelligence work on the open record. Built by intelligence officers. Compliant with ICD 203 and 206. Every claim traceable to its source.
Tradecraft
IC standards, encoded. ICD 203 and 206 are the analytic standards the U.S. intelligence community requires of itself. They are encoded in the pipeline. Every judgment carries a confidence level. Every line of reasoning survives review.
Tempo
Same day. Most questions return in hours, not weeks. Pipelines run thousands of sources at once, across dozens of languages, using human-led, multi-model workflows on a cadence that catches windows before they close.
// THE PIPELINE
One intake. Six engines.
Each engine is built for a different kind of intelligence question. All six work on open-source primary material across dozens of languages. The Intake routes you to the right one.
// HOW IT WORKS / 5 STEPS
01 / 05
Frame.
Start with a plain-language question. The Intake converts it into a PIOT-framed (Population, Intervention, Outcome, Timeframe) intelligence question and asks a short follow-up if anything is ambiguous.
Frame.
Start with a plain-language question. The Intake converts it into a PIOT-framed intelligence question.
Select.
The Watch recommends the best-fit analytic engine and explains why.
Decompose.
Each engine breaks the question into the work required to answer it.
Develop.
Human-led, AI-enabled workflows gather sources, build datasets, and fill information gaps.
Deliver.
The final report is structured for review and decision, with sourced judgments and confidence levels.
// THE STANDARD
Most AI tools ship answers. We ship defensibility.
ICD 203 governs analytic standards: source citation, explicit confidence levels, alternative hypotheses, acknowledgment of uncertainty, clean separation of fact from judgment. ICD 206 governs sourcing. Every claim has to be traceable to the document or observation that produced it. These standards exist because senior consumers will not act on analysis they cannot push back on.
Most general-purpose AI tools can't meet this standard because they weren't built for it. They optimize for fluency, not defensibility. They can't tell you which sentence rests on which source. They collapse complex questions into one-shot prompts. They hide their reasoning. They are, in the technical sense, black boxes.
The Watch was built the other way. Every report is structured to survive that review. The tradecraft is the foundation of the system. The people who built this product spent their careers being held to that standard, and they would not ship something that abandoned it.
The same standard. Different missions.
The Watch operates at the standard set by people who do this work for governments. The standard holds when the buyer is a corporation.
// THE BENCH
Tradecraft is built by practitioners.
The Watch was built by former senior intelligence officers, military officers, OSINT practitioners, and senior national security officials with service across the Intelligence Community. Their experience defines the standard the product now encodes: disciplined framing, rigorous sourcing, transparent reasoning, confidence levels, assumptions, alternatives, and review. This is human-led, AI-enabled, intelligence-grade analysis grounded in real tradecraft.
Meet the team