Generic Advice Is Not Enough
“Prepare an emergency kit.” “Know your evacuation route.” “Monitor official warnings.”
These are good instructions. They are also the same for a coastal household, a family beside a river, an older person living alone, a tenant on an upper floor, and a farmer managing a dry season.
The problem is not that general advice is wrong. The problem is that it stops before the decision becomes personal.
A household usually needs answers such as:
- Does this warning apply to our sitio, street, or building?
- Is the likely danger flooding, wind, storm surge, heat, drought, or a mix of hazards?
- What should we do today, before transport becomes difficult?
- Which family member needs medicine, water, mobility support, or an early check-in?
- What is the official local instruction, and who can we contact if the plan fails?
Preparedness becomes real when a warning is connected to a place, a household, and an action that can still be taken in time.
What Is Hyperlocal Preparedness AI?
Hyperlocal preparedness AI is a decision-support system that combines trusted hazard information with local context. It does not predict the future for a family. It helps people ask better questions and prepare for conditions that may affect their specific area.
For example, the system may bring together:
- Official forecasts, warnings, and local advisories
- Maps of rivers, coastlines, steep slopes, roads, shelters, and service facilities
- Household details people choose to provide, such as number of residents, medical needs, transport options, water storage, and pets or livestock
- Community information, such as barangay contacts, evacuation sites, check-in procedures, and local supply points
- Past reports that reveal which roads flood first, which areas lose power, or where signal coverage is poor
It should state the basis for its answer. A family should be able to see whether a recommendation came from an official warning, a local plan, a submitted household detail, or a map-based estimate. If information is missing or old, the system should say so.
This is not a replacement for PAGASA, an LGU, the NDRRMC, or emergency responders. It is a way to make official information easier to apply before a crisis becomes urgent.
Why Place Changes the Right Action
Two families in the same municipality can need different plans. One may live near a flood-prone creek. Another may be exposed to strong winds on higher ground. A third may have no safe transport if an evacuation notice arrives late.
This is why a map alone is not enough, and neither is a generic chatbot. A useful answer needs both spatial facts and household context.
| Question | Context needed | Useful output |
|---|---|---|
| Should we prepare to leave? | Official alert, hazard zone, shelter access, household mobility | A clear checklist and the official local instruction |
| What should we pack first? | Family size, medicines, infant or elder needs, expected disruption | A ranked list that can be checked and updated |
| Who needs a call today? | Voluntary care network, contact permissions, risk indicators | A private check-in list for authorised community workers |
| What can our block do now? | Local roles, supply points, road and communication constraints | Shared actions with named owners and deadlines |
UNDP’s community resilience guidance treats resilience as broader than one hazard or one government programme. It includes disaster risk, climate adaptation, health risks, conflict prevention, and ecosystem management.
Move From Response to Early Action
Most disaster systems are judged by what happens after an event. Preparedness starts earlier, when a forecast or warning gives people time to reduce harm.
In the Philippines, community-led anticipatory work has used forecast triggers and local mapping to support action before a typhoon makes landfall. UNDRR describes work in Bicol, Eastern Visayas, and Caraga where forecasting, pre-agreed triggers, local alerts, and participatory mapping were used to guide earlier preparation.
A tool should not wait until a family is already in danger to explain what an alert means. It should help them prepare when they still have choices:
- Check the official warning and its update time.
- Identify which households are affected by the stated hazard.
- Suggest simple actions that fit the available time.
- Confirm any local evacuation or safety instruction.
- Record whether assistance, transport, medicine, or a welfare check is needed.
An AI system can organise this information and create a clear plan. It cannot decide to ignore an evacuation order, overrule emergency officials, or turn an uncertain forecast into false certainty.
A Household Plan Should Reflect Real Constraints
Households do not prepare with the same money, time, space, or support. A plan that assumes a car, a stocked pantry, continuous mobile data, or a nearby relative will fail many people.
The first version of a hyperlocal preparedness profile can stay small. It may ask only for information that changes the plan:
- Approximate location or barangay, with the option not to store an exact address
- Household size and who may need added support
- Key medicines, mobility, communication, or power needs
- Water, food, lighting, and charging capacity
- Transport choices and trusted contacts
- Hazards the household has experienced before
The system should ask why it needs each field, who can view it, and how long it will keep it. No one should have to disclose sensitive details just to read general advice.
For instance, an older adult who relies on refrigerated medicine may need a different power and transport plan from a nearby household with the same flood exposure. A farmer may need a drought plan that includes water storage, soil moisture practices, crop timing, and local agricultural support.
Preparedness Is a Community Activity
A family plan is valuable, but it has limits. Shelters, roads, water points, health services, warnings, and communications are shared systems. A household cannot solve a blocked bridge or an inaccessible evacuation centre alone.
PAGASA describes flood preparedness as a community effort and advises families to know their local warning system. It also points to basic household steps such as keeping emergency supplies ready and storing belongings above expected flood levels.
The community layer of a preparedness system should help people coordinate without exposing private information. It can support:
- A shared checklist for barangay volunteers and household leaders
- Confirmed official advisories, with date and source shown
- Voluntary welfare-check requests
- Supply and transport gaps reported in aggregate
- Accessible formats and translated instructions
- Post-event notes about what worked and what did not
It must not publish a list of vulnerable residents, exact addresses, or medical details. A person should choose what is shared, with whom, and for what purpose.
What the System Needs to Do Well
Start with trusted sources
Official forecasts, local disaster plans, public hazard maps, and verified emergency contacts should be the starting point. Every recommendation should display the source and the time it was last checked.
Separate facts, estimates, and advice
“PAGASA issued this warning” is a fact. “Your household may be affected because it is near this mapped flood zone” is an estimate. “Charge phones and move supplies now” is advice. The system should label the difference clearly.
Keep a narrow task boundary
The assistant can explain, organise, prompt, and help draft a plan. It should not make life-and-death decisions, promise that an area is safe, or substitute for emergency services.
Design for weak connectivity
Preparedness cannot depend on an always-on chat screen. Important plans should be printable, downloadable, readable on low bandwidth, and available in plain language. A family should know the next step even when the network drops.
Include people who are often missed
Early-warning systems need to work for older people, persons with disabilities, people with low literacy, informal settlers, remote communities, and people who do not use smartphones. UNDRR calls for early warning and early action that is tailored, gender-responsive, and disability-inclusive.
A Practical System Pattern
A first implementation does not need to cover every hazard or every household. Start with one place, one hazard, and one decision window.
For example, a pilot for flooding in a selected barangay could work like this:
- Gather local knowledge. Meet with residents, barangay officials, health workers, and responders. Document known flood paths, safe routes, signal gaps, and practical limits.
- Connect official information. Use PAGASA advisories and the local DRRM plan as the source of truth. Do not rely on social posts as primary warning data.
- Build voluntary household profiles. Collect only plan-changing information, then give people a way to edit or delete it.
- Create action cards. Turn common situations into brief, ordered instructions: before heavy rain, when the river reaches a stated level, when evacuation is advised, and after returning home.
- Set escalation rules. Make clear when the assistant should point users to emergency contacts, official channels, or a trained local coordinator.
- Run a drill. Test the plan with real users before storm season. Observe whether people understand the advice, can complete it, and know whom to contact.
Test the Plan Against Real Behaviour
A working demo is not evidence that a preparedness system works. The test is whether a person can use it under pressure.
Measure questions such as:
- Did users understand which official warning applied to them?
- Could they find their next action within one minute?
- Did the system point to a current and correct local contact?
- Did users know what information was certain, estimated, or missing?
- Were people with limited data, mobility, or literacy able to use the plan?
- Did community partners find the information useful without receiving unnecessary personal data?
After a drill or event, update the local knowledge. Maybe a route was blocked sooner than maps showed. Maybe instructions used language people did not understand. Maybe a water point had no actual supply. A preparedness system gets better through these corrections, not through confident wording.
A 90-Day Starting Plan
Days 1 to 30: Choose the problem. Select one hazard, one community, and one decision that needs better support. Map trusted sources, decision owners, and information gaps.
Days 31 to 60: Build the smallest useful service. Create sourced answers, household action cards, consent controls, and an escalation path to official help.
Days 61 to 90: Run a controlled drill. Invite a small, representative group of households and local workers. Test access, clarity, privacy, and usefulness. Fix failures before expanding.
For Prixial and Alima AI, the opportunity is not to make another general emergency chatbot. It is to help households turn local hazard information into practical plans for food, water, energy, utilities, and care, while keeping personal data within clear boundaries.
Frequently Asked Questions
Can AI predict whether my house will be damaged?
No. It can combine available maps, forecasts, and household details to explain possible exposure, but it cannot guarantee an outcome. Use official warnings and local authorities for safety instructions.
Does this replace a barangay disaster plan?
No. It should make official and community plans easier for residents to understand and follow. Barangay officials and disaster responders remain responsible for official decisions and response.
What information should a household avoid sharing?
Avoid collecting exact location, medical, financial, or identity information unless it clearly changes the plan and the person has given informed consent. Store the minimum needed, restrict access, and set a deletion period.
Can this work without internet access?
The assistant itself may need connectivity, but the most important outputs should not. Provide printable action cards, saved plans, SMS-friendly instructions, and local contact details.
What is the first use case to build?
Choose a common local hazard with a clear action window. Flood readiness, heat planning, drought preparation, and typhoon readiness are better starting points than a broad all-disasters assistant.
What This Means
Generic disaster advice helps people begin. Hyperlocal preparedness helps them decide what to do next.
The goal is not an AI that sounds certain. It is a tool that gives families and communities timely, sourced, location-aware guidance, shows its limits, protects private information, and makes it easier to act before a hazard becomes a crisis.