Emergency management’s AI challenge isn’t technology — it’s implementation, new reports say

August 5, 2026 - 10:31

New reports from the Markle Foundation's AI for Disasters and Emergencies Initiative suggest that artificial intelligence could significantly improve how emergency management agencies prepare for and respond to crises. But the studies also make one thing clear: the biggest obstacle is not the technology itself. It is getting agencies to actually use it.

The reports point to several practical barriers. Governance tops the list. Many agencies do not have clear rules for when AI can be used, who is accountable for its decisions, or how to handle privacy concerns. Without those guardrails, even a well-built tool can sit on a shelf.

Procurement is another sticking point. Government buying processes are slow and rigid, often designed for hardware or construction, not for software that needs regular updates. By the time a contract is signed, the AI model may already be outdated.

Then there is the skills gap. Emergency managers are trained to coordinate shelters, evacuations and supply chains, not to debug machine learning pipelines. The reports note that agencies need data scientists and AI specialists, but they struggle to compete with private sector salaries.

Still, the potential payoff is real. AI can help forecast flood paths, route ambulances around traffic, or scan social media for early signs of a developing crisis. The Markle initiative calls for pilot programs, shared playbooks and closer ties between tech developers and emergency planners.

The takeaway is simple: the algorithms work. The challenge is building the human and institutional systems around them. Until that happens, the promise of AI in disasters will remain just that, a promise.