In late February and early March 2024, Haiti rapidly devolved into a near-failed state. As one colleague put it, “I want my people to be first in line for rebooking flights, not the back of the line.” This requires speed to action, efficient communications and accelerated discovery of an evolving crisis.

Instability has been a challenge in Haiti going back decades. The country’s infrastructure and political institutions have failed to bounce back after devastating natural disasters in 2010, 2011, 2016 and again in 2021. Since the assassination of Haiti’s president in July 2021, the country has fallen victim to rampant gang violence and 80 percent of the capital, Port-au-Prince, is controlled by gangs. Despite these challenges, businesses, NGOs and government personnel have continued to operate and travel in the country.

Content Provided By OnSolveDuring this crisis, OnSolve’s Global Risk and Intelligence Services provided complimentary overwatch support for Travel Management Company (TMC) customers and their travelers in Haiti. The AI-powered platform provided real-time tracking, verification and reporting as gang warfare broke out on February 29, 2024. OnSolve Watch Center analysts provided additional support as a human operational layer (aka human-in-the-loop) by verifying the severity of the threat and notifying TMC customers according to customized SOPs.

Early morning reports on February 29 indicated that planes were damaged by gunfire and injuries were reported in the airport's main lounge. This was quickly followed by reports that Sunrise Airways was suspending operations. At the same time of these reports, flight radar, Google and similar flight tracking applications showed that most flights were expected to depart, essentially as “normal."  We had to make the judgment call that other companies would suspend operations — and we were right.  This is the work of the analyst: to take multiple points of data (parsed with AI from human-validated sources) and make a judgment call.  Absent Haiti's police or security apparatus, which was clearly diminished, this airport would close and people would be trapped. 

Risk Events

  • In the early morning of February 29, 2024, source data reported heavy gunfire and barricades in Port-au-Prince. Geolocated data indicated that the gunfire drew closer to the Toussaint Louverture International Airport (MTPP, PAP) as the early morning progressed. Despite morning reports of airport assets being hit by gunfire, flight tracking applications reported flight plans as “normal."
  • By the afternoon of February 29, 2024, Sunrise Airways suspended flight operations nationwide due to violence in Port-au-Prince. Other airlines were still scheduled to depart that day and the airport remained open and operational.
  • On March 2, 2024, multi-source reporting confirmed coordinated gang warfare targeting police forces, government facilities and the country’s largest prison, resulting in thousands of inmates escaping.
  • On March 3, 2024, a state of emergency and curfew was declared.
  • By March 4, 2024, all flights were canceled, and the airport was closed after gangs attempted to seize control.
  • On March 10, 2024, U.S. Marines evacuated non-essential embassy personnel.
  • On March 11, 2024, Haiti’s prime minister announced his agreement to resign.

Implications for Risk Managers

What does the future hold for operations in Haiti? How do we improve risk intelligence capacity? How should we think about AI and human analysts?

What’s next in Haiti?

  1. The prime minister has tendered his resignation following pressure from other Caribbean leaders and the announcement by Kenya to postpone plans to send police forces until after a new government is formed in Haiti. Prime Minister Ariel Henry remains head of state until after a transitional council has been officially established. Gang leaders will continue applying pressure until the full removal of the prime minister.
  2. Gang violence will continue at an elevated level, impacting core infrastructure and services until the UN peacekeeping mission is fully deployed to Haiti and a new government is formed.
  3. In the following weeks, gun violence will be pervasive in urban centers as police and rival gangs battle for control and efforts are made to detain escaped inmates.
  4. Evacuation operations continue out of Toussaint Louverture International Airport (MTPP, PAP) and other secured runways and helipads. However, recent reports of commercial aircraft being shot at suggest that air evacuations will become severely limited in the coming days.
  5. The long-term outlook of the country is directly tied to assistance from the international community. After foundational security is reestablished in Haiti, international organizations and government support must find a way to return. OnSolve is proud to support travelers and organizations doing good in high-risk environments. Technology solutions and partnerships will strengthen their resilience to continue this important work.

How is it possible to monitor risk everywhere, 24/7?

  1. The rapidly developing crisis in Haiti highlights the value AI brings to risk management. Our ability to manage travel risk is directly proportional to our risk intelligence capacity. It’s time-intensive and cost-prohibitive for analysts to process the sheer volume of risk data available today. And yet, it can feel like we need all the information we can get to make faster, more informed and more effective decisions. It would take hundreds of people to process the content and alerts OnSolve’s AI produces every day.
  2. In a previous blog, I outlined the importance of determining proximity, severity and impact when assessing risk. We can use these same principles to refine risk intelligence gathering. AI solutions must have intuitive filters based on organizational data, travel itineraries and customizable alerts by security teams. Invest in customizable alerts by finding the right blend of pushed versus pulled intelligence.
  3. Risk intelligence is about speed, accuracy and relevance. Past technology solutions have felt like different levers we push based on organizational capacity, risk tolerance and needs. Some pushed intelligence felt fast while some felt more accurate. Some felt overly noisy but comprehensive, while others felt overly generic and unactionable. The beauty of AI and its future is that we can have the best in speed, accuracy and relevance. This requires a human partnership committed to optimizing an AI solution.

Is AI replacing the human analyst? What is the harmonious model?

  1. AI can’t replace the analyst, but it might replace much of the tedious work analysts (particularly junior analysts) do today. The OnSolve Watch Center is a proof-of-concept for how analysts and the engine of AI-generated intelligence work together. We dogfood the platform’s reporting, provide tight feedback with AI engineers and are continually optimizing the filters and alerts based on lessons learned as power users. This empowers us to guide customer adoption of AI risk intelligence and provide additional overwatch as needed.
  2. I don’t think we’re ready for AI to make tough judgment calls. Telling travelers that Toussaint Louverture International Airport is going to close while multiple sources demonstrated ongoing operations is not an easy call. Travelers in Haiti are more comfortable operating in high-risk environments and less inclined to proactively leave the country. Every travel risk manager can attest to how difficult it is to convince an employee to change their plans based on risk. This requires empathy, compassion, tact and flexibility — not exactly how we’d characterize AI today. Travelers and employees don’t want AI telling them to evacuate — it must come from a human in the loop.

The mission can feel daunting and the path forward unclear. If you’d like to continue this discussion, provide feedback or are looking for assistance, OnSolve is here to help.

(Information cut-off date 1000 PT, March 12, 2024)