Agencies are buying AI for emergency response. Many have built systems that’re not good at predicting things and that is a big problem. Before an agency asks if its AI is good enough it should ask if its data is good enough. Most of the time the answer is no.
Being “AI-ready” is not something you can buy. It is a property of your data.
Every country has a special number that people can call in an emergency. For example people in India and Europe call 112, people in the US and Canada call 911. People in the UK call 999. Having a special number is just the beginning. What really matters is what happens after someone calls for help. Most of that is decided before the call is even made, based on how the data is set up.
Every emergency response system is judged on what it takes to get help to someone who needs it. In India, the Emergency Response Support System has been able to get help to people. For example, in Haryana the average time it takes to get help to someone has gone from over 16 minutes to under 10 minutes. The next big improvement is supposed to be AI, which can spot problems before they happen and send help around to them.
Imagine it is raining hard and there is a storm in Mumbai. The emergency call center is getting a lot of calls about flooding, accidents, and power lines being down. If the system has data, it can tell the dispatcher which roads are likely to be flooded and which route to send the ambulance. If the system does not have good data it is like every storm is the first storm.
The problem is that most agencies do not have data. It is not because their algorithms are bad. It is because their systems were not designed to remember things from the past.
The Data Model Decides What is Possible
You cannot just add AI to a system later. If the database was designed to show what is happening now it will not be able to predict what will happen in the future. Whether or not a system can predict things is decided when it is first being built, usually by someone who does not realize how important that decision is.
A system that only shows what is happening now will get rid of the one thing that predictive AI needs: information from the past. The important decisions about a system are made early on by engineers who are trying to get the system up and running. They are not thinking about what an analyst will need two years later. As a result, the system is not able to predict things and by the time that is realized the information from the past is already gone.
Lineage: The “Why” Behind Every Change
The cost of a system with no memory is high. Imagine a system that only stores the state of things: who is assigned to what, the status of an asset and where it is. It looks like it has all the information it needs. Every time something is updated the old information is gone. Two years later when someone wants to know how response times have changed the answer is not available. It was never recorded. The system has been getting rid of its memory all along.
To be able to predict things a system needs to preserve its history. That is not enough. It also needs to know why things changed and when. For example if someone calls to report a crime and then a minute later a traffic system logs a speeding vehicle, the system should be able to link the two events. If the system only stores the current state of things those two events will stay separate.
Why This Is a CXO Question, Not an IT One?
This is not a problem for emergency response systems. It is a problem for all organizations. The cost of forgetting is not always as obvious as it is in emergency response. It is still there. AI does not create the past. If the database got rid of yesterdays information no model can get it back. The answers the model gives will sound just as confident as if it had the information.
Being “AI-ready” is not something you can buy. It is a property of your data. It is decided years before you train a model on it. In practice, this means deciding early on which information will be important to analysts in the future and setting up the system to store that information. For agencies that are buying systems, it means including those expectations in the contract than hoping the system will have them already.
The Decision That Matters Most
The decision that matters most is the one that is made when the system is being built. The agencies that do well will be the ones whose database was designed to remember things from the past. Prediction is a decision, and it is decided early on. In emergency response that decision can mean the difference between life and death. It is worth making it a careful decision, not a footnote, in a contract.
Authored by Ramanandham Konduru, Executive Director, Product Development, Octave, a company that provides mission-critical software that empowers organizations to make informed decisions across every stage of the asset lifecycle. Konduru specializes in building high-performing organizations, driving operational excellence, and delivering complex technology initiatives across Public Safety, CRM, and Healthcare domains. Any opinions expressed in this article are strictly those of the author.
