HUMANITARIAN AI · GOVERNANCE · LOCAL LEADERSHIP · 2026

Humanitarian AI in 2026: The Case for Governance and Local Leadership

Q: What are the three essential actions to close the governance gap?

Three critical actions are: organizational AI policies, mandatory disclosure of AI use in decisions affecting communities, and contextual bias assessment for tools trained on OECD data.

Individual AI tool adoption across the humanitarian sector continues to outpace organizational readiness, creating a critical governance gap. Based on research reaching over 4,200 respondents from 144 countries, 93% of humanitarian professionals now use AI tools, with 75% using them daily or weekly. Yet only 23% work in organizations with formal AI policies. By the time institutions catch up, AI use has already contributed to decisions affecting vulnerable communities—without audit trails, bias checks, or accountability mechanisms. This case study demonstrates that AI in humanitarian and international development is not primarily a technology challenge, but a governance, leadership, and systems change challenge.

Focus: Humanitarian AI governance & local leadership Research: 4,200+ respondents, 144+ countries Key contexts: Kenya, Sudan, Bangladesh, Nigeria, Philippines, Somalia Lead: Humanitarian Leadership Academy (HLA) & Data Friendly Space (DFS)

Read original: How are humanitarians using AI in 2026? The case for governance and local leadership

What is this case study about?

This case study synthesizes findings from the second phase of humanitarian AI research conducted by the Humanitarian Leadership Academy (HLA) in partnership with Data Friendly Space (DFS). Written by Ka Man Parkinson, HLA's Communications Lead, as a personal reflection, it draws on data from 4,200 survey responses collected across two waves (2025-2026), reaching more than 2,700 individuals through online sessions, and thousands more through events, podcasts, and social media engagement.

The research reveals a sharpening paradox: individual AI adoption is accelerating rapidly while organizational readiness remains largely static. This gap—combined with the fact that the highest usage is concentrated in crisis-affected regions where local organizations have the least governance support—leads Parkinson to argue that humanitarian AI in 2026 is primarily a governance and protection challenge, not an innovation agenda.

Central argument: AI in international development is not primarily a technology challenge, but a governance, leadership, and systems change challenge. The question is shifting from "should we use AI?" to "how can we use it responsibly?"

Figure 1: The AI adoption paradox — high individual usage versus low organizational readiness

The core problem: Adoption without governance

Most humanitarian professionals use AI tools like ChatGPT, Claude, or CoPilot for work. Yet few have ever asked whether the LLM outputs were checked for bias. According to the 2025 foundational study by HLA and DFS—the first comprehensive global baseline of its kind, covering 2,539 practitioners across 144 countries—93% of humanitarian professionals use AI tools, with 70% integrating them weekly or daily. However, only 22% work in organizations with a formal AI policy and governance in place.

Across the board, staff adopt tools faster than institutions adopt frameworks. AI becomes embedded in workflows before anyone assesses its specific use cases, risks, data sources, and biases. By the time the institution catches up, AI use has already contributed to decisions affecting communities—without an audit trail, a bias check, or an accountability mechanism.

Figure 2: Local organizations are the highest daily users of AI but have the lowest governance capacity

The research methodology: A two-phase global study

Phase 1 (May - November 2025)

The first comprehensive global study into how humanitarians are using AI reached more than 2,500 respondents from 144 countries. As the research co-lead noted: "we had tapped into a massive underground conversation", signaling huge demand for insights and guidance on navigating AI in humanitarian work.

Phase 2 (January - March 2026)

Building on the first phase, the team focused on rapid data collection, community engagement, and mobilization through surveys and digital platforms:

January 2026 pulse survey: Tracked shifts in adoption patterns and attitudes, reaching more than 1,700 respondents from 120+ countries
Three global online sessions (January-March) in partnership with NetHope and at Humanitarian Networks and Partnerships Weeks (HNPW), generating 1,700+ registrations
Research briefing note released in March in English, French, and Spanish, alongside an updated interactive data dashboard

Total reach: 4,200 survey responses | 2,700+ individuals through online sessions | Thousands more through events, podcasts, social media, and sector media including Devex.

Key geographic insight: Over 80% of respondents are from the Global South or Majority World—even stronger representation than the 2025 baseline study (75%). The highest growth and most intensive daily usage are concentrated in regions with acute humanitarian needs, including Kenya, Sudan, and Bangladesh.

Key findings: The AI adoption paradox

The crises and upheaval of 2025 appear to have deepened the paradox rather than resolved it: rising individual conviction set against largely static organisational readiness. AI adoption is bottom-up, not top-down. Organizations are lagging behind practitioners.

Accelerating individual adoption

93% of humanitarian professionals use AI tools (2025 baseline)
75% use AI daily or weekly (+5% since 2025)
65% say AI has improved operational efficiency (+18%)
54% feel AI has supported better decision-making (+16%)
30% are now using custom-built AI agents—a term that did not appear in the 2025 survey

Static organizational readiness

Only 9% work in organizations where AI is widely integrated (+1%)
Only 23% have a formal AI policy (+1%)
Only 3% consider themselves expert AI users—essentially unchanged

Figure 3: The growing gap between individual AI adoption and organizational readiness

A senior leader in operations at a local NGO in Somalia said:
"AI has strong potential to improve humanitarian work by supporting data collection, faster decision-making, and better targeting of assistance. However, more training and access are needed to ensure effective and responsible use, especially at local level."

A programme manager at a local NGO in the Philippines said:
"Since our work deals with people with complex situations, I think what AI can help in our organisation is that it can help us analyse scientifically, but it cannot replace human interface. So ultimately, we will adopt with caution."

Closing the governance gap: Three essential actions

To address the governance gap and ensure responsible AI adoption, practitioners and sector leaders have identified three critical actions:

1. Organizational AI policies

Define acceptable and prohibited use cases, especially for decisions affecting vulnerable populations. Policies must distinguish between low-risk administrative tasks and high-risk decisions involving targeting, eligibility, or protection.

2. Mandatory disclosure

Require disclosure when AI tools inform assessments, targeting, or reporting that drives funding or programming decisions. Transparency enables accountability and allows affected communities to understand how decisions affecting them were made.

3. Contextual bias assessment

Recognize that commercial AI tools trained predominantly on OECD country data perform differently in crisis and low-resource environments. Organizations must conduct bias assessments specific to their operational contexts.

The humanitarian and development sectors already have strong accountability frameworks: do no harm, accountability to affected populations, minimum standards. AI adoption should be governed by the same principles, not treated as an exception.

Figure 4: Many organizations are stuck between individual experimentation and pilot projects, lacking the governance structures to move to full integration

2026: Governance comes into sharp focus

At the end of the first phase of research (November 2025), no clear consensus was emerging on sectoral priorities. Five months on, AI literacy and governance are crystallising as critical priorities—conversations, data, and sectoral movements point to increased convergence around these as foundational challenges.

With the rapid diffusion of AI across the ecosystem—driven by accessible LLMs and now agentic AI—and limited movement on organisational AI governance, the evidence points to humanitarian AI as a governance and protection challenge, rather than primarily as an innovation agenda.

A practitioner reflected on the governance gap:
"The fact that AI policy is at a slow pace compared to the growth in use within the humanitarian sector is concerning. Governance frameworks provide the context in which AI can truly serve the public good."

Another expert noted:
"Clear policies don't stop innovation—they give people the confidence to start using AI in a responsible and open way."

Local leadership: The real issue is power, not technology

In both 2025 and 2026, what really emerged was the creative and resourceful applications and approaches of local actors in the Global South—a finding documented by multiple research efforts: "the Global South leads innovation." More AI use cases and tools are being developed locally, particularly specialised language models and domain-specific solutions tailored to regional needs.

As one practitioner working on local AI leadership framed it:
"When we talk about local leadership in humanitarian AI, we often focus on access to technology, but from my perspective, the real issue is power, not technology."

Another expert added:
"Let's not overestimate the risks and underestimate the opportunities. Local organisations in Nigeria, Lebanon, Syria, Sudan, Kenya, Rwanda are giving us very good examples of how to leverage these tools. The main important thing is not to stand in the way of local organisations and local leaders. AI tools and AI in general could be either the best or the worst thing that could ever happen to humanity and to what we do. And localising AI could take us to the best-case scenario."

A practitioner from Uganda summarised the need:
"We need an artificial intelligence that speaks the language of the donor and the language of the village where I come from. We need an AI that is good for all of us."

Active work must be done in the AI space to counter the risk that 'humanitarian AI' becomes shorthand for large-scale technical deployments by well-resourced international actors, particularly when current adoption patterns are globally distributed, bottom-up, and often without governance or organisational support.

Key takeaways for evaluators and M&E practitioners

Individual adoption outpaces organizational readiness

93% of humanitarians use AI tools, but only 23% have formal policies—a governance gap that creates protection risks for vulnerable populations. AI adoption is bottom-up, not top-down.

Local organizations lead adoption but lack governance

Local organizations are the highest daily users of AI, yet only 13% have formal AI policies (vs. 39% in UN agencies). Innovation is bottom-up and local; governance is top-down and uneven.

AI is a governance challenge, not just innovation

The evidence points to humanitarian AI as a governance and protection challenge. Risks include data misuse, ethical concerns, bias, harm, and environmental impact.

Three actions can close the governance gap

Organizational AI policies, mandatory disclosure of AI use in decisions affecting communities, and contextual bias assessment for tools trained on OECD data.

The real issue is power, not technology

Local leadership in AI requires shifting power, not just providing access to technology. Local actors should be at the table as co-designers, not testers.

AI agents are emerging rapidly

30% of respondents are now using custom-built AI agents—a term that did not appear in 2025. This rapid evolution demands proactive governance.

The right to say no must be preserved

Universal individual uptake does not mean universal organizational adoption is inevitable. 10% of organizations have no intention of adopting AI. Intentional non-adoption is a valid choice.

AI = organizational change, not just IT

"AI is not an IT initiative. It is an organisational change initiative." It requires leadership, culture shift, training, and intentional implementation.

Frequently asked questions

What is the humanitarian AI paradox?

The humanitarian AI paradox is the sharpening gap between rising individual AI adoption (93% use AI tools, 75% daily or weekly) and largely static organizational readiness (only 23% have formal AI policies). AI adoption is bottom-up, not top-down, creating a governance vacuum where individual usage continues without organizational guardrails.

What are the three essential actions to close the governance gap?

Three critical actions are: (1) Organizational AI policies that define acceptable and prohibited use cases, especially for vulnerable populations; (2) Mandatory disclosure when AI tools inform assessments, targeting, or reporting that drives decisions; (3) Contextual bias assessment recognizing that commercial AI tools trained on OECD data perform differently in crisis environments.

How is AI adoption distributed geographically?

AI adoption is not following a Global North-to-South diffusion pattern. The highest growth and most intensive daily usage are concentrated in regions with acute humanitarian needs, including Kenya, Sudan, and Bangladesh. Over 80% of survey respondents are from the Global South, and the Global South leads innovation in AI use cases.

What is the governance gap between local and international organizations?

Only 13% of local organizations have formal AI policies compared to 39% in UN agencies, despite local organizations being the highest daily users of AI. The governance gap falls hardest on those already working with fewer resources, despite them being at the forefront of AI adoption and innovation.

What is the main takeaway for humanitarian and development organizations?

AI in international development is not primarily a technology challenge, but a governance, leadership, and systems change challenge. Organizations need to develop fit-for-purpose AI policies, invest in digital leadership, ensure local actors are empowered as co-designers, and recognize that AI adoption is not obligatory—intentional non-adoption is a valid choice.

Main reference and original source

This case study is based on the personal reflection and research findings by Ka Man Parkinson and the Humanitarian Leadership Academy, as documented in:
Parkinson, K. M. (2026, April 14). Opinion | How are humanitarians using AI in 2026? The case for governance and local leadership. Humanitarian Leadership Academy.
https://www.humanitarianleadershipacademy.org/resources/opinion-how-are-humanitarians-using-ai-in-2026-the-case-for-governance-and-local-leadership/

Additional sources:
Humanitarian Leadership Academy & Data Friendly Space (2025). First comprehensive global baseline study of AI adoption in the humanitarian sector. 2,539 practitioners, 144 countries.
Humanitarian Leadership Academy & Data Friendly Space (2026). January 2026 pulse survey. 1,700+ respondents, 120+ countries.

Interpretation and analysis by EvalCommunity: The insights, lessons, and expert commentary presented in this article are offered by EvalCommunity as an independent interpretation of the research findings for the M&E and humanitarian evaluation community. They are based on the original opinion piece by Ka Man Parkinson and the Humanitarian Leadership Academy's research outputs. EvalCommunity does not claim ownership of the original research or findings. We encourage readers to consult the original source for the complete context.

Resources for further learning

Advance your skills in AI governance and responsible evaluation

Learn how to develop fit-for-purpose AI policies, lead organizational change, and center local actors in AI-enhanced M&E practice. Access courses, templates, and expert guidance.

EvalCommunity Services AI in M&E Course

Humanitarian AI – Governance – Local Leadership 2026