Understanding where damage occurred 

In the first hours after a disaster, responders are trying to determine the extent of the damage, who needs help and what supplies should be sent where. Every hour spent gathering information can mean an hour not spent delivering aid. 

Rivas reached out to Microsoft’s AI for Good team shortly after the earthquake hit June 24, and by the next morning he had a map analyzing the damage. The free, open-source HASTE tool requires only a few hours of someone training it on a specific situation — analyzing satellite images captured before and after a disaster and identifying damaged buildings — before it’s able to use those patterns to assess vast territories that would otherwise take people days or weeks to comb through. 

Catholic Relief Services, which provides aid in about 95 countries around the globe, had been experimenting with AI, says Nora Lindstrom, who leads the organization’s global digital programming team from Helsinki. But producing an assessment like the one HASTE provided overnight would have taken her group two weeks, requiring analysts, imagery and a far more manual process, she says. That’s too long to be of help in the immediate aftermath of a disaster. 

“We have limited resources, and we’re trying to use those as efficiently as possible to maximize the good that we can do and the people that we can help,” Lindstrom says. “And speed is really key to that.” 

But accuracy is just as important.  

“If you have fast garbage, that’s actually worse,” says Kevin White, a senior director at Microsoft’s AI for Good Lab. “AI allows you to get high precision, fast — and that’s the holy grail.” 

Teams might be preparing to spend 12 hours traveling through difficult terrain to one town based on incomplete reports, Rivas says, only to later discover that another community in the opposite direction suffered far greater damage. 

“It’s about going fast and having certainty about your decision,” he says. 

Catholic Relief Services used HASTE’s AI-assisted analysis to help responders with local partner Caritas Venezuela identify the hardest-hit areas and prioritize aid after the June 2026 earthquakes in Venezuela. (Photos courtesy of Caritas Venezuela)

Turning information into action 

For the United Nations Office for the Coordination of Humanitarian Affairs (OCHA), quickly turning information into action can make the difference between reaching survivors and arriving too late. 

“You’re saving lives by providing this data to the team who is doing search and rescue, whose job is to bring out people who are under the rubble and might be alive,” says Fawad Hussain Syed, a humanitarian affairs officer with OCHA in Geneva. “What this data is doing is basically getting to those teams in time so they can go with their dogs and machinery and find people who are living and bring them out.” 

After disasters, information often arrives in fragments through messages, news reports, social-media posts and scattered field observations that can be difficult to verify, Syed says. Satellite imagery with analysis provides critical evidence. 

“It translates into a response plan,” he says.   

OCHA received HASTE’s Venezuela damage analysis and rapidly distributed it through a humanitarian network used by thousands of people from governments, nonprofits, local authorities and emergency-response teams. That helped organizations begin building a shared understanding of where damage appeared most severe so they could estimate needs, plan operations and communicate the scale of the disaster to partners and donors. 

Syed says that’s “very, very crucial” for coordinating search-and-rescue responses, in particular, as teams arrive from all over the world to help — sometimes 10 to 15 of them, depending on the disaster. 

“They land at the airport and the first thing they want to know is where to go,” he says. “This is exactly what this data then provides.” 

Understanding the full scope of a disaster 

Not long after the earthquake, responders faced a new challenge. 

Groups were getting overwhelmed with imagery and analyses from multiple sources “flooding the system,” says Zachary Arno, a data scientist with OCHA. He quickly built a web viewer, hosted on Azure, that brought HASTE assessments and eight other disaster-mapping products together in one place so organizations could compare them and build a more complete picture of what they were all showing about conditions on the ground.

The Microsoft data arrived with detailed metadata and confidence rankings, Arno says, making it easier to incorporate into that broader analytical picture. 

“Speed and communication and transparency helped shape the viewer,” he says. 

Even with abundant satellite imagery, understanding the full scope of a disaster is difficult.  

AI assistance is most useful when there’s an uneven pattern of damage, with some communities devastated while others remain largely untouched, says Caleb Robinson, a principal research scientist for AI for Good. Every disaster leaves a different visual footprint, Robinson says, and buildings themselves can vary dramatically from one region to another. Machine learning can quickly identify the hallmarks of damage in a specific event and then apply those patterns across thousands of images. 

Leonardo Milano, OCHA’s data science team lead, says having an automated tool “doing analysis over much larger areas so you’re sure you’re not missing areas that may be particularly affected” is important.  

“It’s the combination of speed and scale that is really the added value,” Milano says. 

Recovery begins with estimates 

As wildfires tore through the Spokane, Washington, area in August, Tristan Allen, who coordinates with businesses and critical infrastructure for the state’s Emergency Management Division, reached out to the AI for Good team to see “what the art of the possible was to identify numbers and structures within the fire line.”  

Firefighters were still battling blazes engulfing hundreds of homes, and officials knew formal assessments would come later, Allen says. But HASTE’s early look at the scale of the destruction unfolding helped emergency managers begin estimating how many displaced households they’d need to prepare to support. 

“If you’re looking at thousands of structures lost, you’re going to surge more resources earlier,” Allen says. “In other disasters, when you don’t have any idea or you just have rough guesses, it takes a while to get that process up and running.”

Building local capacity

HASTE has quickly become Microsoft’s primary platform for disaster-response assistance. During major emergencies, the AI for Good team often activates alongside humanitarian partners, working with Planet Labs and other imagery providers to help process images, generate assessments and distribute information. But HASTE is open-source, allowing organizations to adapt it for their own systems and needs. 

Rivas quickly realized that aid workers wouldn’t be able to reach crumpled communities if roads and bridges were equally damaged, so he adapted the HASTE analysis to evaluate routes, identify potential obstacles and suggest alternatives. 

Because HASTE is open-source, Catholic Relief Services adapted the technology to support the specific needs of local partner Caritas Venezuela after the June 2026 earthquakes in Venezuela, from helping responders navigate damaged routes to mapping aid distribution for donors. (Photos courtesy of Caritas Venezuela)

He’s also using HASTE to show accountability, creating a map of tarp distribution to help donors visualize how their funds had been used to provide shelter in the most damaged neighborhoods.

“What’s been great about the HASTE approach is that it’s open-source, and it’s really sought to co-create with us as well,” Lindstrom says. “People who know the context on the ground are best placed to respond, and so strengthening their ability to do so and to use these new technologies in this way is super powerful.” 

That capability is especially important for smaller disasters that might not attract global attention, Rivas says, such as recent fires in Bolivia and floods in Peru, Paraguay and Ecuador. Responders are already getting more comfortable and quicker using AI assessments, he and Syed say — with HASTE maps informing decisions within an hour, not a day, of the Colombia earthquake in August.

Imagining new uses 

The Spokane fires sparked another idea for how AI could help after disasters. 

Survivors need to know where they can go to get water, food, shelter and medicine — or even just where they can charge their phone to stay in touch with loved ones, says Joseph Porcelli, who coordinates with public agencies for social media app Nextdoor. Reaching them is one of the biggest challenges governments and humanitarian organizations face, he says. 

HASTE used before-and-after satellite imagery from the Spokane, Washington, wildfires in August 2026 to identify damaged structures across the region, helping emergency managers estimate the scale of the destruction and get an early view of how many households might need support. (HASTE screenshots courtesy of Microsoft AI for Good)

When Porcelli saw the HASTE assessment from Spokane, he ran the info through Nextdoor’s database and realized most of the affected households were members. He’s now working on a program to connect the platform with HASTE data so local governments with official accounts can target more specific geographic areas to help in future emergencies. 

Not only that, aid organizations are starting to use it as a tool for preparedness, not just response. Catholic Relief Services is using the damage-assessment data from Hurricane Melissa, which struck Jamaica last October, to help plan shelters in areas that could be vulnerable in future storms, Rivas says.  

There is “a fast-evolving culture of consuming this data,” says Rivas, who has worked in humanitarian response for about 15 years and is part of a global initiative that helps local teams better prepare for and respond to disasters. 

AI for Good’s “goal is to provide first responders with the most accurate and up-to-date information we can, as quickly as possible,” Microsoft’s White says. “We’re encouraging people to use these tools, to download them, to build this capability for themselves, too.”

This story was originally published on Sept. 22, 2026.

Top image: Caritas Venezuela, a Catholic Relief Services partner, assesses damage following the June 2026 earthquakes in Venezuela. (Photo courtesy of Caritas Venezuela)
Scrolling images: Aerial photos from the 2026 Venezuela earthquakes (HASTE screenshots courtesy of Microsoft AI for Good)

Susanna Ray writes about AI and technology, with stories that show its real‑world impact and examine how innovation is reshaping work, business and society. She previously reported for Bloomberg News and other major international news organizations in the U.S. and abroad, covering beats ranging from politics and government to business and aviation. Follow her work on Microsoft Source.