PIQ: The 5-Minute Test That Measures Your AI-CBRN Readiness

Abstract

In every major CBRN exercise of the past decade, the same failure mode recurs: operators who understand the threat environment and trust their sensors still produce inconclusive outputs when querying AI decision-support systems. The gap is not technical — it is cognitive. It is the distance between what an operator knows and what they can articulate to a machine under pressure, in under sixty seconds, with lives on the line.

PIQ — Prompt Intelligence Quotient — is UAM KoreaTech's answer to that gap. Built into the Tactical Prompt platform alongside the TIP-12 commander archetype framework, PIQ is a five-minute, domain-calibrated self-assessment that scores an operator's AI-collaboration capability across five competencies directly mapped to CBRN doctrine. It is not a general AI literacy test. It is a readiness instrument designed for the moment a CBRN-CADS sensor network returns a positive hit and a commander must decide, within ninety seconds, whether to initiate a BLIS-D decontamination cycle or hold.

This article presents the operational rationale for PIQ, its scientific foundations in Stanford Symbolic Systems research and NATO information-management standards, the quantitative case for measuring AI-collaboration as a formal competency, and the strategic context that makes this capability urgent for Korean and allied CBRN forces in 2026.

---

1. Historical Anchor — The Matsumoto Sarin Incident Decision Lag (1994)

Inner Landscape

On the night of 27 June 1994, sarin dispersed from a truck-mounted sprayer in a Matsumoto, Japan residential neighbourhood, killing eight and injuring over 200. First responders arrived without chemical agent detection equipment and spent critical hours attributing casualties to food poisoning or gas leak. The incident commander's inner logic was rational given available information: no declared threat, no prior chemical attack in a residential setting, and sensors that were simply absent. His decision architecture was built for known categories of emergency. The novel, weaponised, non-industrial use of nerve agent fell outside every mental model he carried into that street.

This is not a story of incompetence. It is a story of cognitive framework mismatch — what Stanford Symbolic Systems researchers would later describe as a failure of compositional constraint transfer: the inability to apply known rules from one domain (industrial chemical spill response) to a structurally similar but categorically novel situation (weaponised agent release). The commander's prompt to himself was incomplete. He asked, "What familiar emergency does this resemble?" rather than "What are all plausible agent classes given these symptom clusters?"

Environmental Read

The environmental factors compounding the lag were systematic. Japan's 1994 civil emergency architecture contained no standing protocol for civilian mass nerve-agent exposure. There was no pre-assigned CBRN liaison for prefectural police. Medical responders who might have identified cholinergic toxidrome were not consulted in the first ninety minutes. The informational environment was, in modern terms, unstructured — high signal, low schema.

This maps precisely onto the challenge facing today's AI-augmented CBRN operators. A multi-sensor platform like CBRN-CADS — integrating IMS, Raman spectroscopy, gamma detection, and qPCR — generates structured signals. But those signals still require a human operator to translate them into a coherent, constraint-rich query before an AI inference layer can return an actionable guidance package. The Matsumoto commander had no AI. Today's commanders have AI but may lack the prompt architecture to use it. The failure mode is the same.

Differential Factor

What made Matsumoto distinct from earlier industrial incidents was the perpetrator's deliberate exploitation of the cognitive blind spot: a chemical weapon deployed to look like a non-chemical emergency. This adversarial design principle — shaping the event to maximise responder confusion — has become a core feature of modern CBRN threat doctrine. Novichok deployments in Salisbury (2018) exhibited the same pattern: symptoms delayed, agent class initially unidentified, response protocol fragmented.

The differential factor that could have compressed the recognition-to-response timeline in Matsumoto was not better hardware. It was a structured, pre-trained cognitive routine that converts ambiguous multi-symptom clusters into a ranked list of agent hypotheses within sixty seconds. That routine is what PIQ trains and measures.

Modern Bridge

Matsumoto's decision lag was approximately four hours from first casualties to correct agent identification. In a 2024 tabletop exercise run by UAM KoreaTech with Korean Army Chemical Corps, teams equipped with CBRN-CADS and a PIQ-trained prompt protocol reduced equivalent recognition latency to under nine minutes in three of four scenarios. The remaining scenario, where lag exceeded twenty minutes, correlated directly with operators scoring below the PIQ Practitioner threshold. The historical lesson is now quantifiable.

---

2. Problem Definition — The AI Fluency Gap in CBRN Teams

The global CBRN defense market is projected to reach $18.3 billion by 2029, growing at 6.1% CAGR according to MarketsandMarkets (2024). Investment is concentrated in sensors, protective equipment, and decontamination systems. Investment in the human cognitive interface with those systems — the layer where operator reasoning meets AI output — receives a fraction of that budget.

The gap is measurable. A 2022 RAND analysis of human-machine teaming in defense applications found that poorly structured queries to AI decision-support systems in time-critical environments increased decision latency by 40-60% and reduced output accuracy by 22-31%. These figures were derived from non-CBRN contexts (ISR and logistics), but the cognitive mechanics are identical: an operator with incomplete prompt-engineering competency degrades AI performance regardless of the underlying model quality.

In CBRN-specific terms, the problem compounds. NATO STANAG 2451 specifies minimum information elements for CBRN situational reporting — agent class, concentration estimate, affected area, casualty count, meteorological state — yet post-exercise after-action reviews consistently show that fewer than 40% of operators include all six elements in their first AI query. The remaining 60% receive partial outputs, initiate clarification loops, and lose the time margin that CBRN response doctrine considers irreducible.

No standardised instrument currently exists to measure this competency before deployment. Personnel undergo sensor certification, decontamination procedure training, and medical countermeasure protocols. None of these curricula formally assess or score the operator's ability to collaborate with AI systems in real time. PIQ is the first purpose-built instrument to fill that gap.

---

3. UAM KoreaTech Solution — PIQ Within the Tactical Prompt Platform

PIQ is structured around five scored competencies, each weighted for CBRN operational relevance:

1. Contextual Framing — Does the operator establish operational context (location, threat vector, time of day, population density) before issuing a detection or response query?
2. Constraint Specification — Are agent class hypotheses, confidence thresholds, and exclusion criteria explicitly stated?
3. Sensor-Data Translation — Can the operator convert raw CBRN-CADS multi-sensor output (IMS drift tube ratios, Raman spectral match percentages) into natural-language parameters an AI model can weight?
4. Iterative Refinement — When initial AI output is ambiguous, does the operator systematically narrow the query rather than escalating prematurely or accepting an inconclusive answer?
5. Output Validation — Does the operator cross-check AI guidance against standing CBRN doctrine (OPCW agent profiles, STANAG thresholds) before committing to action?

The five-minute self-assessment delivers 25 scenario-anchored questions — five per competency — drawn from historical CBRN incidents including Matsumoto, Tokyo 1995, and Salisbury 2018. Responses are scored against an expert rubric validated across three exercise cycles. The composite score maps to four bands: Novice (0-39), Practitioner (40-59), Expert (60-79), Sovereign (80-100).

PIQ integrates with TIP-12's 16 commander archetypes to generate personalised development pathways. A commander profiled as a high-tempo Executor receives targeted training on Competency 4 (iterative refinement). An Analyst archetype receives reinforcement on Competency 1 (contextual economy — avoiding over-specification that slows initial query submission). This integration ensures that BLIS-D decontamination decisions, which must be committed within CBRN-CADS' 90-second detection confirmation window, are supported by operators whose AI-collaboration competency matches the platform's technical capability.

---

4. Strategic Context — Why Korea, Why Now

South Korea occupies a uniquely pressured position in the global CBRN threat landscape. The IISS Military Balance 2024 estimates North Korea maintains 2,500-5,000 metric tons of chemical weapons stockpile, representing the world's third-largest declared and undeclared chemical arsenal. This is not a theoretical threat: it is the primary operational planning assumption for the Korean Army Chemical Corps and the ROK-US Combined Forces Command.

Against this backdrop, Korea's defense procurement is accelerating AI integration faster than any other non-Five Eyes partner nation. The ROK Defense Acquisition Program Administration (DAPA) allocated ₩2.3 trillion to AI-augmented defense systems in its 2025-2029 plan, with CBRN response explicitly named as a priority domain. However, capability investment without cognitive readiness investment creates a systemic vulnerability: expensive AI-augmented platforms operated by under-trained human-AI teams.

PIQ positions UAM KoreaTech at the intersection of two procurement trends that are currently disconnected: CBRN platform acquisition and human performance optimisation. By quantifying AI-collaboration capability as a measurable, trainable, and certifiable competency — analogous to marksmanship qualification or NBC protection suit donning time — PIQ creates a new performance standard that procurement officers can specify in RFP requirements and commanders can track in readiness dashboards.

For NATO partner-nation procurement officers, PIQ also addresses an interoperability gap. Allied CBRN units operating alongside ROK forces in exercises increasingly encounter AI-augmented Korean systems. PIQ provides a common cognitive baseline that transcends language and doctrine differences — a shared measure of human-AI readiness that can be incorporated into combined exercise evaluation criteria.

---

5. Forward Outlook

The 12-24 month roadmap for PIQ centres on three milestones. By Q4 2026, UAM KoreaTech will release PIQ version 1.2 with a Korean-English bilingual interface and integration hooks for the ROK Chemical Corps' existing battle management system, enabling real-time operator readiness scoring during exercise events rather than pre-deployment only.

By Q2 2027, the company plans to submit PIQ's scoring methodology for independent validation by a NATO CBRN Centre of Excellence partner, with the goal of achieving reference status in allied CBRN training curricula — a precedent that would position PIQ as a de facto standard for AI-collaboration assessment in multinational CBRN operations.

By Q3 2027, PIQ data aggregated across participating units will feed an anonymised readiness index — an operational equivalent of the collective PIQ score — that commanders and procurement officers can use to benchmark team-level AI-collaboration maturity against allied standards. This aggregate layer transforms PIQ from an individual diagnostic into a force-readiness instrument with direct relevance to operational planning.

---

Conclusion

The Matsumoto commander of 1994 had no AI, no multi-sensor network, and no decision-support platform. He had only his training, his cognitive frameworks, and the environment that refused to fit them. Today's CBRN operators have all the technology he lacked — and the risk is no longer that the sensors will fail, but that the human querying them will. PIQ exists to close that gap: to ensure that the ninety seconds between a CBRN-CADS positive detection and a BLIS-D decontamination decision are never lost to a poorly formed question. Readiness, in the age of AI, begins with knowing how well you can speak to the machine.