What CAD Actually Does — and What It Doesn't
How a 911 call becomes an incident, how units are recommended and dispatched, how apparatus status and location are tracked, and why so many other department systems depend on CAD.
Read
Data, AI, and Administration
Fire and EMS agencies already collect enormous amounts of information. The challenge is getting trustworthy data out of disconnected systems, turning it into information people can actually use, and applying automation or AI only where it improves a defined problem. The same is true of the systems that run the organization rather than the incident. Inventory, scheduling, budgeting and planning are where most of a department's administrative time goes, and where most of its data is created in the first place. They belong here for that reason: none of them is a separate technology problem, and all of them fail the same way, by producing numbers nobody can defend.
What technology needs to solve
Documentation takes too long and data quality suffers
Incident and patient-care documentation competes with operational readiness and other responsibilities. Repetitive entry, poorly designed workflows and disconnected systems increase workload while inconsistent documentation reduces the value of the information later.
We collect data nobody uses
Years of CAD, RMS, ePCR, staffing and response data may exist while deployment and operational decisions are still made primarily from anecdote.
Our data is trapped in separate systems
CAD knows one part of the incident. RMS knows another. ePCR contains patient care. GIS knows location. Staffing systems know who was available. Useful analysis often requires connecting information that was never designed to work together.
We do not know what equipment we have, where it is, or when it expires
Time goes on locating equipment before it is needed, service and expiration dates are found late or found by an inspector, and nobody can answer a straightforward question about what the department owns.
Scheduling and overtime consume administrative time
Filling a shift takes an officer off other work, callback order is argued about after the fact, and the department cannot say what overtime is actually being spent on until the year is over.
We cannot defend a budget request with our own numbers
A request built on last year's request is argued on opinion, and a governing body that cannot see the basis for a number treats every number in the submission the same way.
We cannot show what the department will need in three years
Station location, staffing and apparatus decisions are made one at a time as each becomes urgent, and each one is argued from scratch.
What good should look like
- Departments retain practical access to their own information.
- Data can be exported in usable formats.
- Definitions are consistent.
- Dashboards answer operational questions rather than show every available field.
- AI output can be verified.
- Automations reduce duplicate work.
- Integrations have clear ownership.
- Privacy and cybersecurity are addressed.
- Analytics distinguish correlation from causation.
- Human judgment remains responsible for consequential decisions.
- Administrative systems produce a record, not just a screen somebody looked at.
- The department can get its own roster, inventory and financial data out in a usable format.
- A number in a budget request can be traced back to the record it came from.
- Systems that collect information about individual employees are agreed before they are deployed.
Learning
Learning LabRecords Management and the Move to NERIS
What fire department records systems do, how incident data moves from CAD into reporting systems, what changed with the transition from NFIRS to NERIS, and why fire departments should understand their data even when a vendor submits it for them.
Read
Cyber Incident Response: When Systems Go Down
What the first hours look like when CAD, records or email are gone, who decides what comes back first, and how a department keeps answering calls while it happens.
Read
Reading Your Own Incident Data
Learn how to read response-time distributions, workload, unit utilisation, concurrency, incident types and other operational measures without being misled by a single average.
Read
Dashboards People Actually Use
A dashboard should help someone make a decision. Start with the question and the audience rather than filling a screen with every field available in the database.
Read
AI for the Fire Service: What It Is Good At
Artificial intelligence is useful for some tasks and unreliable for others. Learn where current tools can save time, identify patterns, summarise information and assist people — and where confident output should not be trusted.
Read
AI Governance and Policy
Establish the rules before widespread adoption. Define acceptable uses, sensitive information, human responsibility, records requirements, procurement expectations, and what must happen when AI is wrong.
Read
Still in development
Innovation LabAI-assisted incident command decision support
Systems that analyze radio traffic, unit status, sensor information, location, benchmarks or other incident data and attempt to identify important changes that deserve the incident commander's attention.
Why it matters
The near-term opportunity is not an autonomous incident commander. It is a second set of eyes that may help a human recognize something important under high cognitive workload. The risk is false confidence, missed events, unnecessary alerts, and automation bias. Ask what exact decision the system is supporting, what information it monitors, how often it is wrong, and what its false positives and false negatives look like. Ask whether the IC can understand why an alert occurred, what happens when the data is incomplete, and who remains responsible for the decision.
Peer experience
Mutual AidHas anyone integrated ESO with First Due?
Running ESO for EMS records and looking at First Due for fire. Both vendors say it is supported. Looking for someone who has actually run it in production.
How is your NERIS migration actually going?
Vendor says we are ready. Our data says otherwise. Curious whether anyone has completed a clean migration and what they had to fix first.
Recommended research
Government, standards and peer-reviewed sources on this problem. Every address here was checked before it was published.
Research
AI Risk Management Framework
A voluntary framework for identifying and managing the risks of an AI system.
Why it matters
A neutral structure for the questions to ask a vendor selling AI, and the basis of a defensible department policy.
Research
Generative AI Profile
The companion profile covering risks specific to generative AI.
Why it matters
Directly relevant to assisted report writing. Confabulation is a named risk here, which is why a generated narrative is reviewed before anyone signs it.
Official resource
NEMSIS Public Reports
National EMS data reported publicly.
Why it matters
A reference point for what your own numbers look like against the national picture.
Hub guide
Learning Lab — Data and AI
Related guide on the Hub.
Hub guide
Records Management and the Move to NERIS
Related guide on the Hub.
Hub guide
API, Interface, Integration: What Vendors Mean
Related guide on the Hub.
Hub guide
ePCR, NEMSIS, and the EMS Data Chain
Related guide on the Hub.
How these sources are labeled
- Research
- Academic or laboratory work.
- Official resource
- A government or standards body. Neutral, no commercial interest.
- Hub guide
- Written by the Hub.
Nothing here is paid placement. Links are chosen because they are the clearest explanation we could find, and labeled so you know who wrote them.