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

Emergency managers see artificial intelligence as a promising tool to ease growing workloads, but the biggest hurdle to adoption isn’t a lack of technology, it’s building the governance, procurement processes and workforce skills needed to deploy it responsibly, according to a trio of reports published Tuesday.

The reports were produced by the AI for Disasters and Emergencies Initiative of the Markle Foundation, a nonprofit focused on national security, in partnership with Aspen Digital, part of the Aspen Institute, a nonprofit that strives to improve public policy, and RAND, a nonprofit think tank.

The research arrives as state and local emergency management agencies face increasingly frequent and costly disasters while operating with limited staff and budgets. Researchers found that AI can help emergency managers process information faster, automate repetitive administrative work and improve situational awareness during disasters.

The reports identified more than 1,100 AI-enabled products from more than 700 companies. While vendors have concentrated heavily on disaster response capabilities, such as hazard monitoring and damage assessment, the research found fewer tools tailored for emergency preparedness, recovery and cross-agency coordination.

“Understanding that AI is designed as an augmentation tool, not a replacement tool,” Jeremy Greenberg, a senior adviser for Aspen Digital, said in an interview. “Then understanding that any of the adoption of of artificial intelligence has to be based in trusting how you’re using it, where the human factors are still coming into play and how it’s being applied.”

The reports’ findings build on Aspen Digital’s AIDE Initiative, which launched earlier this year to identify practical, responsible uses for AI in emergency management. Emergency managers participating in Aspen Digital’s workshops identified 40 potential AI use cases, with the strongest demand centered on situational awareness, operational coordination and public information.

Greenberg said AI could also reduce the time spent analyzing weather data, drafting emergency plans, communicating with the public and completing administrative work such as grant applications and after-action reports — since the tools are well suited for repetitive, information-heavy tasks that currently consume valuable staff time.

“There are certain things where it’s like if you can offload that data crunching, then turn the human to doing the analysis of the product that comes out of it. I think that’s where we also get a win,” he said.

Greenberg said procurement remains one of the biggest obstacles for government because many AI vendors rely on token-based pricing models that do not align with traditional government purchasing processes.

The reports found that fragmented products, opaque pricing models, integration challenges and privacy concerns as persistent barriers to adoption. Agencies also wield increasingly complex technology stacks, in which an AI capability may depend on several other products.

“In one sense, it gives a benefit to the buyer if you want to do a la carte and you only want to purchase a specialized capability. But at the same time, the bundling that needs to happen is coming from a variety of different companies,” Greenberg explained. “Stacking them appropriately is the internal focus, then figuring out how to purchase them externally in bundles is the external approach.”

Greenberg recommended first establishing governance frameworks, workforce training and procurement guidance that helps enable responsible adoption within their budgets.

“Understanding what you’re trying to achieve, what you have available, and what the barriers are to to bringing these systems in house — not just in emergency management organizations, but enterprise-wide across public safety, public works,” Greenberg said. “With better governance comes better adoption.”

To help agencies navigate those challenges, the AIDE Initiative plans to launch AI education programs, publish a practical AI adoption playbook, establish pilot projects with emergency management organizations and convene government, industry and academic leaders at an AI summit in Austin, Texas, this September.

“True adoption has to be baselined in trust. This is not this is not hardware that you can touch and feel. It is software inside of a a computer that you are interacting with,” Greenberg said. “You have to understand what you’re using it for, acknowledge what the information is that’s coming out of it, and how you’re applying it in emergency management setting.”

Despite rapid advances in AI, Greenberg said emergency management will always need to depend on human judgment.

“The decisions about life and safety in that immediate threat environment and into the immediate response, you absolutely still have to have a human making critical decisions,” he said. “But where we’re seeing augmentation, the technology is providing better analysis, so the human can make a better decision.”