AI in crisis management: situation reports, crisis communication and aftercare, with the mayor in charge
Short answer
When an incident escalates, the safety region takes over the response and the mayor holds supreme command. The municipality delivers population care: shelter, communication, aftercare, registration. AI can summarise incoming reports and messages into a situation picture, draft messages in plain language and several languages, and prepare the evaluation afterwards. What AI does not do: decide on evacuation, or communicate without a person having seen the message.
The structure
The Dutch Safety Regions Act places disaster response with the safety region and supreme command with the mayor. On escalation the municipality’s teams move: the municipal policy team, the duty officer for population care, the communications adviser and the teams for shelter, registration and aftercare. Everything moves fast, with incomplete information, under pressure from media and social media. Germany and France organise this through their civil protection structures and communal emergency plans.
Confidential information is in it from the start: casualty data, addresses, business data, and sometimes information from police and prosecutors.
Where AI helps
| Task | What AI does | What the crisis team does |
|---|---|---|
| Situation picture | Summarises reports, messages and log lines into a timeline with sources | Interprets and decides |
| Crisis communication | Draft messages for website, social media and public-warning text, in plain language and the languages of the neighbourhood | Checks and sends |
| Environment analysis | Bundles what is being said online about the incident and flags rumours | Decides whether and how to respond |
| Residents’ questions | Answers practical questions from the approved messages | Handles the rest |
| Evaluation | Reconstructs the timeline from logs and messages for the evaluation report | Draws the lessons |
See Plain-language letters with AI: in a crisis a message everyone understands counts double.
Where it goes wrong
- Speed over control. An invented address or a wrong instruction in a crisis message cannot be undone. Every message is read by a person before it goes out, however busy it is.
- Manipulation. During an incident messages pour in that are not all reliable. A summary must not turn rumours into facts; have AI attach the source and the certainty.
- Outage. An AI environment that depends on an external service fails when the connection drops. A node in your own building or in a domestic data centre keeps working.
- Information sharing. Casualty data and police information must not go into a tool outside the authority. See Information security and AI in local government.
- Exercises. An assistant used only in a crisis is misused in the crisis. Include it in the exercises.
How werqly fits in
Crisis management is regional: several municipalities share the safety region and the plans. A shared node for the region, or nodes per municipality with the same sources, fits that; see AI in joint arrangements between municipalities. werqly runs in your own building or in werqly Cloud in a Dutch data centre, keeps working without an external service, and logs who had which situation picture and which message produced. See AI for local government.
Frequently asked questions
Questions on this topic
May AI send crisis messages?
Draft them, yes; send them, no. A person reads every message before it goes out. An error in a crisis message, a wrong address or a wrong instruction, cannot be undone.
Can AI build a situation picture from reports and social media?
Yes, as a summary with source and certainty per element. Rumours must be marked as rumours. Interpreting and deciding stay with the crisis team and the mayor.
What if the internet connection drops during a crisis?
An AI environment that depends on an external service then fails. A node in your own building keeps working; a node in a domestic data centre stays reachable as long as your own connection holds. Include this in the continuity plan.
May casualty data go into an AI tool?
Only in an environment of the authority or the safety region, with access per role and logging. Casualty registration and police information never belong in a public tool.
Sources
- Dutch Safety Regions Act (Wet veiligheidsregio’s)
- National Coordinator for Counterterrorism and Security (NCTV)
This article is general information, not legal advice. Rules and guidance change; check the sources listed or ask your lawyer or DPO if in doubt.