1 of 21

Demystifying

AI

A Leader’s Guide to AI-Powered Decision-Making

2 of 21

Your Guides

Umair Rehman, PhD

Assistant Professor Computer Science -

Western University, Department of Computer Science,- Faculty of Science

Scott D. Ramey, MHSc

Assistant Chief, PhD Student - Halifax Regional Fire & Emergency, University of Waterloo - Faculty of Engineering

3 of 21

The 3 AM Problem: A Hazmat Scenario

Let's start with a scenario. It's 3 AM and there is a major hazmat discharge at a local industrial operation.

You need:

Chemical information

Sensors

PPE

Company contacts

Facility floor plans/shut offs

Environmental and waterway information

Notifications

Evacuations

Decision Support

"This is a massive information problem under extreme time pressure. Right now, the IC is juggling a physical ERG, trying to get dispatch to pull up maps for evauaction zone, and calling for more resources. Contacting Canuetec. Conducting research with various different software applications.

Now, imagine by the time the IC arrives, the system has already identified the probable chemical from a business provided pre-plan. It's automatically running a plume model based on live weather data. It's flagging a seniors home downwind for evacuation, and providing floor plans pulled directly from the HazReady Community Safety app where that business owner uploaded their details. While the IC makes decisions, the system is providing automated accountability tracking of every firefighter on scene, it's monitoring their biometrics through their wearables, breathing air levels, assessing technician for stress levels based on their biometrics, and it's even pulling in live air-quality readings from a sensor deployed by a robot. This isn't the future; this is exactly what we are building now.

Supporting Research:

Human Factors - Situational Awareness: This scenario highlights the challenge of achieving situational awareness, defined by Endsley (1995) as perceiving critical elements in the environment, understanding their meaning, and projecting their future status.

AI tools can directly support all three levels. (Reference: Endsley, M. R. (1995). Toward a theory of situation awareness in dynamic systems. Human Factors, 37(1), 32-64.)

4 of 21

ADMINISTRATION/PREPAREDNESS: Policies & guidelines are often outdated and seldom used for real-time decisions

Our Shared Reality: Growing Complexity

INCIDENT MANAGEMENT: Accountability and firefighter safety are still manual and tragically error-prone.

EMERGENCY MANAGEMENT: ICs and EOCs struggle with situational awareness and multi-agency coordination.

QUALITY IMPROVEMENT/LEARNING: After-Action Reviews are inconsistent and delayed, failing to drive real improvement.

And the incident is just the start. The administrative burden is crushing Chiefs. We're defending budgets, handling grievances, and proving compliance. We need tools that reduce this burden, not add to it.

What if policies and guidelines can be uploaded for incident context and recommendations, but also, the system flags them for review and renewal, or changes to NFPA or OH&S standards?

Is it time to move beyond velcro accountability tags? Why are there no real successful digital accountability systems? HUMAN FACTORS - we are solving this.

The government of Canada released a white paper in 2012 on Interoperability and Common Operating Picture for emergency services and emergency management, nobody has tackled this problem. Meet HazReady!

Our very own wildfire in Halifax that evacuated 16,000 people and destroyed over 200 homes, took almost a year to complete the PIA. – We propose this can be done in minutes. And provide recommendations based on standards, policies, guidelines, and actual incident data from CAD, RMS, ICS forms, and more.

Supporting Research:

Human Factors - Cognitive Load: The administrative burdens and ever-changing policies contribute to high cognitive load on leadership, even outside of active incidents.

This can impact strategic decision-making and organizational management. (Reference: Sweller, J. (1988). Cognitive load during problem solving: Effects on learning. Cognitive Science, 12(2), 257-285.)

5 of 21

6 of 21

The AI Engine: A Look Under the Hood

Vector Database (The Smart Library): Finds conceptually related information, not just keywords.

LLM (The Language Engine): Processes language to understand context and generate responses.

MoE (The Committee of Experts): Routes questions to specialist models for faster, more accurate answers.

To understand how this works, let's look under the hood. You'll hear a lot of terms, but they boil down to a few key ideas.

First, the Vector Database and talk about the embedding models.. This is the heart of the 'smart search.' We use an embedding model to turn all of your SOPs into a complex map of meaning. This is what allows you to ask a question in plain English and get the conceptually right answer, even if the keywords don't match.

Next, the LLM, or Large Language Model. This is the powerful engine that understands and generates language. You've likely interacted with these already.

More advanced systems use a Mixture of Experts, or MoE model. Think of this like having a digital committee of specialists. When you ask a Hazmat question, the system doesn't bother the high-angle rescue expert. It intelligently routes the query to the most relevant 'expert' model, making the system faster and more efficient.

Supporting Research:

Vector Embeddings: (Reference: Mikolov, T., et al. (2013). Efficient estimation of word representations in vector space. arXiv preprint arXiv:1301.3781.)

Mixture of Experts: (Reference: Shazeer, N., et al. (2017). Outrageously large neural networks: The sparsely-gated mixture-of-experts layer. arXiv preprint arXiv:1701.06538.)

7 of 21

The Right Tool for the Job:

Responsible AI vs. Consumer AI

Consumer AI (Public Web Tools)

Knowledge Source: The entire public internet.

Risk: Can "hallucinate" (invent facts). Your data is often used for training.

Responsible AI (The RAG Approach)

Knowledge Source: A closed loop based on your documents.

Benefit: Grounded in fact, private, and secure. It provides citations for every answer.

This brings us to the most critical point: safety and trust. Using a consumer AI for public safety is a non-starter. They are designed to be creative, which means they can make things up, and your data becomes part of their system.

RAG: Your AI's "Trusted Advisor": Think of RAG (Retrieval-Augmented Generation) as the system that makes HazReady a safe and reliable AI, unlike general internet chatbots.

No Hallucinations, Only Facts: Public AIs can "make things up." HazReady's RAG prevents this by forcing the AI to only use information from your department's verified documents (SOPs, policies, pre-plans). It's always grounded in your truth, and keeps your data isolated and protected.

Your Data, Your Control: All your documents are stored in a secure, Canadian-hosted Vector Database. This database is completely separate and private.

Transparency & Accountability: Every answer is traceable. The RAG system provides citations back to your original source document, giving you full auditability and confidence in the information.

The "Walled Garden": It creates a secure "walled garden" of knowledge. It becomes an expert only on your department's specific rules and data, ensuring relevance and maximum security.

Why This Matters: RAG is the technical cornerstone that ensures a secure, trustworthy, and compliant AI partner for emergency services, protecting your data and preventing dangerous misinformation.

Supporting Research:

Retrieval-Augmented Generation: (Reference: Lewis, P., et al. (2020).

Retrieval-augmented generation for knowledge-intensive NLP tasks. Advances in Neural Information Processing Systems, 33, 9459-9474.)

Trustworthy AI: (Reference: NIST AI Risk Management Framework (AI 100-1), National Institute of Standards and Technology, 2023.)

8 of 21

Built for Trust: A Foundation of Security

Canadian Data Residency: Your data stays in Canada.

Immutable Records: SecureLedger™️ provides blockchain-secured, tamper-proof incident records.

Enterprise SSO and role-based access control.

Government-Grade Security: Built to SOC 2 compliance standards

9 of 21

The Vision: A Connected Intelligence Ecosystem

"The neural link between your responders, your EOC, and your knowledge."

10 of 21

The Brain: Fire Service Intelligence

A secure, Canadian built AI engine that acts as the "brain of the system."

Manages your documents, policies, guidelines and builds training plans.

Helps draft memos and respond to grievances with your Collective Agreement.

Tracks resources, peoples, qualifications

and expertise.

Embedded across all modules as the emergency service specific AI agent.

Synapse™️Intelligence is the core AI engine—the 'brain of the system'. It's the RAG system I described. We feed it systemeic knowledge (NFPA, CSA, CPSE, ISO, IFSTA, etc.) You feed it all your policies, SOPs, training manuals, staffing, assets, equipmetn, fleet, memo’s and even collective agreements. It can then help you draft memos, training plans, all-hazards plans, HIRA’s, emergency management plans or even initial grievance responses based on your own documents. Don’t have an RMS? It also tracks qualifications and expiries, embedding this intelligence across all modules.

Supporting Research:

Knowledge Management Systems: Synapse™️ functions as an advanced knowledge management system, designed to capture, distribute, and effectively apply organizational knowledge, which is critical for high-reliability organizations. (Reference: Nonaka, I. (1994).

A dynamic theory of organizational knowledge creation. Organization Science, 5(1), 14-37.)

11 of 21

What does this look like for the end user?

12 of 21

On Scene: Command & Responder

Predictive Safety Index to protect responders before a Mayday.

Incident Command Dashboard with automatic accountability & biometric monitoring.

Primary General CoAI-driven Incident Action Plans (IAPs), logs and ICS forms.

Agentic AI: Autonomous AI agents to monitor tasks.

HazReady Responder is the next generation of intelligent incident management (Actually, realistically the first). In our Hazmat scenario, it generates an IAP in minutes, pulling data from the Community Safety module. It provides a live dashboard with automatic accountability and biometric monitoring."

This is where we move towards Agentic AI. An AI Agent is like a dedicated digital assistant you can give a goal to. Instead of you constantly checking the screen, you can task an agent: 'Monitor Sector B's air supply and alert me when the first cylinder hits 15 minutes remaining.' The agent then works autonomously in the background to achieve that goal, freeing up your cognitive space to focus on strategy.

Supporting Research:

Intelligent Agents: Research in this field focuses on autonomous entities that can perceive their environment and act to achieve goals, a core concept for reducing human operator workload in complex systems. (Reference: Russell, S. J., & Norvig, P. (2010). Artificial intelligence: A modern approach. Pearson Education.)

13 of 21

14 of 21

15 of 21

In the EOC: Situational Awareness - COP

Provides a real-time common operating picture for the EOC.

AI-assisted Hazard Identification & Risk Assessments (HIRA).

Enables multi-agency interoperability and mutual aid coordination.

Generates AI Post-Incident Analysis and After-Action Reports.

HazReady EM™️ (Emergency Manager) is your AI-powered Emergency Manager. It gives your EOC a real-time common operating picture, allowing them to see all active incidents, coordinate mutual aid, and even use AI to draft internal and public communications. It uses your organizational emergency plan, policies, guidelines, assets, critical infrastructure, personnel, and more to give you real-time incident decision support.

Still on paper ICS forms? Let’s talk about the future of ICS in built by HazReady.

Critically, after the incident, the system collates all data and drafts an After-Action Report, identifying lessons learned and feeding them back into Synapse and a clear action plan for continuous improvement.

Supporting Research:

Common Operating Picture: The platform provides a COP, which is empirically shown to enhance shared situational awareness and improve coordination effectiveness among multi-agency teams during emergency response. (Reference: Wolbers, J., & Boersma, K. (2013). The common operational picture as collective sensemaking. Journal of Contingencies and Crisis Management, 21(4), 186-199.)

16 of 21

In the Community: HazReady Community Safety

Data feeds directly into Responder (tactical pre-plans) and EM (strategic planning).

Businesses create site profiles with hazards, floor plans, contacts, inspections.

Connects the private sector to emergency services.

The biggest challenge with pre-planning is outdated information. HazReady Community Safety solves this. We give businesses a portal to manage their own site profiles with hazards, floor plans, and contacts. This data feeds directly into the Responder and EM platforms, creating a living, breathing pre-plan maintained by the people who know the building best.

How does your system become truly resilient and anti-fragile? It requires extensibility. Imagine you are the EM for Vancouver and there is an earthquake. You need heavy machinery at multiple locations overwhelming your system. The local caterpillar dealer has shared their equipment inventory in the HazReady Community Safety business portal and accepted the state of emergency resource sharing agreement. Now you can mobilize thie heavy machinery right away.

Residential App: FireSmart, Fire Safety Plan, 72 Hour checklist, Fire Prevention Public Education, incident reporting during declared emergency.

Supporting Research:

Community Resilience: This module facilitates public-private partnerships, a key component in building community resilience by leveraging local resources and knowledge for more effective disaster preparedness and response. (Reference: Comfort, L. K., Boin, A., & Demchak, C. C. (Eds.). (2010).

Designing resilience: Preparing for extreme events. University of Pittsburgh Press.)

17 of 21

The Value: Who Benefits?

Enhanced preparedness and admin efficiency.

For

Chiefs

Real-time accountability and decision support.

For Incident Commanders

AI support from planning through post-incident analysis.

For Emergency Managers

Confidence in a system that protects their lives.

For Responders

Greater resilience, awareness and faster recovery.

For

Communities

"A complete Al-powered emergency intelligence ecosystem."

18 of 21

Today - Narrow AI: Highly specialized AI for specific tasks.

Theoretical - AGI (Artificial General Intelligence): AI with human-like cognitive abilities.

Science Fiction - ASI (Artificial Superintelligence): AI that would surpass the brightest human minds.

The AI Horizon: Hype vs. Reality

Finally, let's address the big picture. You'll hear terms like AGI and ASI. It's important to know what they mean.

AGI, or Artificial General Intelligence, is the theoretical goal of an AI that can learn and reason like a human. This is the stuff of research labs, not fire grounds. ASI, or Superintelligence, is the science fiction concept of an AI that vastly exceeds human intellect.

It's crucial to be clear: we are not building science fiction. HazReady is Narrow AI. It is practical, specialized, and built for the sole purpose of making the fire service safer and more effective, today.

Supporting Research:

On Intelligence: (Reference: Legg, S., & Hutter, M. (2007). A collection of definitions of intelligence. Frontiers in Artificial Intelligence and applications, 157, 17.)

Superintelligence Concepts: (Reference: Bostrom, N. (2014). Superintelligence: Paths, Dangers, Strategies. Oxford University Press.)

19 of 21

This Future Must Be Built Together

20 of 21

An Invitation: Become an Innovation Partner

Join a national leadership group on AI for fire services in Canada.

Benefits:

Help shape the future of emergency services technology.

Secure discounted early adopter pricing.

Scan the Code to Join the Innovative Leaders List

21 of 21

Questions & Discussion

"Thank you. We'd now be happy to answer any questions you may have."

The AI-powered emergency intelligence ecosystem that connects responders, emergency operations centres, and communities for safer, smarter, and more resilient operations.

Is your organization HazReady?

gethazready.com